Bing Chat now has ads(twitter.com)
twitter.com
Bing Chat now has ads
https://twitter.com/debarghya_das/status/1640892791923572737
248 comments
I’ve yet to see an example of *GPT be better at search than Google.
I’ve seen a few examples that breathlessly proclaimed that they literally saved lives or saved hours of time, but when I typed the 3 most obvious keywords from their input into Google, the first result always has the answer right there.
I’ve seen a few examples that breathlessly proclaimed that they literally saved lives or saved hours of time, but when I typed the 3 most obvious keywords from their input into Google, the first result always has the answer right there.
I never have a search problem.
I need either information, a service or a product. Or information about a service or a product.
I never need a search, a website or an app. That's never the problem I'm trying to solve.
Search, websites and apps are the tools I have to use (currently) to solve my problems. In the process I have to:
- deal with the fact that the interest of the makers of the app or website do not align with mine
- deal with privacy invasions/ads/notifications
- deal with things being phrased disingenuously
- deal with apps needing to update, website being broken/bandwidth wasted
And most importantly, unless two vendors implemented an integration together, I have to integrate their services myself!
I never want to leave the chat. I want to do all my transactions from the chat. All of them. On my terms. With my prompts.
Whoever facilitates that can take a 5-30% cut on all my online transactions.
Examples:
- book me a highly rated hotel in Berlin around 100 euro p/n most suitable for a staycation for next week Monday to Friday. Make a reservation near that hotel for the highest rated traditional German cuisine restaurant with a price per plate of 20-30 euro
- send every contact ive emailed with from my work email in the last year a Christmas card with a funny but professional pun about the invention of the steam engine. Send it to their work address.
I need either information, a service or a product. Or information about a service or a product.
I never need a search, a website or an app. That's never the problem I'm trying to solve.
Search, websites and apps are the tools I have to use (currently) to solve my problems. In the process I have to:
- deal with the fact that the interest of the makers of the app or website do not align with mine
- deal with privacy invasions/ads/notifications
- deal with things being phrased disingenuously
- deal with apps needing to update, website being broken/bandwidth wasted
And most importantly, unless two vendors implemented an integration together, I have to integrate their services myself!
I never want to leave the chat. I want to do all my transactions from the chat. All of them. On my terms. With my prompts.
Whoever facilitates that can take a 5-30% cut on all my online transactions.
Examples:
- book me a highly rated hotel in Berlin around 100 euro p/n most suitable for a staycation for next week Monday to Friday. Make a reservation near that hotel for the highest rated traditional German cuisine restaurant with a price per plate of 20-30 euro
- send every contact ive emailed with from my work email in the last year a Christmas card with a funny but professional pun about the invention of the steam engine. Send it to their work address.
> - book me a highly rated hotel in Berlin around 100 euro p/n most suitable for a staycation for next week Monday to Friday. Make a reservation near that hotel for the highest rated traditional German cuisine restaurant with a price per plate of 20-30 euro
I imagine this will go about as well as if you chose to purchase an apartment purely on the basis of the real estate agency's description.
I imagine this will go about as well as if you chose to purchase an apartment purely on the basis of the real estate agency's description.
This is what I do now on booking.com. If you have a smarter strategy let me know. But whatever the strategy is: it can be automated.
I also spent a lot more time on the choice where I live than the hotel in going to stay at for a week. Weird.
And although ChatGPT currently doesn't have up to date information about hotels or the ability to book it (a simple booking.com plugin would enable that though), it is smart enough to either pick a hotel with remote working facilities or consider one with coworking spaces nearby.
I also spent a lot more time on the choice where I live than the hotel in going to stay at for a week. Weird.
And although ChatGPT currently doesn't have up to date information about hotels or the ability to book it (a simple booking.com plugin would enable that though), it is smart enough to either pick a hotel with remote working facilities or consider one with coworking spaces nearby.
I don't know if it's a smarter strategy, but you can send a travel agency your list of requirements and they'll send you back a quote. My parents' generation are probably the only people that still use agencies, but it works for them.
It takes longer than a search query, but they don't want the hassle of coordinating logistics, and have some peace of mind that they're getting what they asked for.
It takes longer than a search query, but they don't want the hassle of coordinating logistics, and have some peace of mind that they're getting what they asked for.
So what's the bussiness model of the travel agency?
- they take a cut
- they are neutral because they want repeat bussiness
- they are personalised around your requirements
- they take care of the little details
It's exactly what I'm describing here, except automated. Because at heart their work is analyst work and thats exactly what you can automate with a LLM
- they take a cut
- they are neutral because they want repeat bussiness
- they are personalised around your requirements
- they take care of the little details
It's exactly what I'm describing here, except automated. Because at heart their work is analyst work and thats exactly what you can automate with a LLM
> except automated.
The devil is always in the details. A travel agency builds contacts with airlines, hotels, tour groups etc, so they can have a backup option available if your first preference is not. That requires humans.
An AI system that is simply browsing through the same websites a human would is going to come up with a lot of "no" responses for specific booking requests, or will simply quote higher prices once the booking websites figure out how to detect them.
The devil is always in the details. A travel agency builds contacts with airlines, hotels, tour groups etc, so they can have a backup option available if your first preference is not. That requires humans.
An AI system that is simply browsing through the same websites a human would is going to come up with a lot of "no" responses for specific booking requests, or will simply quote higher prices once the booking websites figure out how to detect them.
I'm not suggesting OpenAI is maintaining those contacts. I'm suggesting booking.com will upload an API integration into the ChatGPT platform. Just like they uploaded an app into the AppStore.
Perhaps they will try to up the price on this platform compared to others. But only until AirBnB comes to compete on this platform too (and uploads their API integration)
It's a platform game.
There was a time when Google tried to index everything and afterwards came a time when everything tried to be indexed by Google. Once you are the default, they have to be there and they will have to compete with others there.
And as long as the chat protects users from losing control over their intent, that's where the users will be and that's where you will need to be offering your products to be viable. Obviously at that point the regulators come knocking..
Assuming there would only be one player. There won't be, but for now only OpenAI is actually aware of the opportunity and Microsoft doesn't even seem to understand it's a platform game now and not an ad game.
Perhaps they will try to up the price on this platform compared to others. But only until AirBnB comes to compete on this platform too (and uploads their API integration)
It's a platform game.
There was a time when Google tried to index everything and afterwards came a time when everything tried to be indexed by Google. Once you are the default, they have to be there and they will have to compete with others there.
And as long as the chat protects users from losing control over their intent, that's where the users will be and that's where you will need to be offering your products to be viable. Obviously at that point the regulators come knocking..
Assuming there would only be one player. There won't be, but for now only OpenAI is actually aware of the opportunity and Microsoft doesn't even seem to understand it's a platform game now and not an ad game.
What makes you think those results would be neural, and not sponsored/SEOed the same way everything is these days? What makes you think the machine works for you and not the advertisers? Also, to this day, LLMs/ANNs are not capable of explaining their outcome, and you won't argue/reason with them. That's a pretty big difference if you ask me.
>What makes you think those results would be neural, and not sponsored/SEOed the same way everything is these days?
Because it would be the less profitable bussiness.
>Also, to this day, LLMs/ANNs are not capable of explaining their outcome, and you won't argue/reason with them
How would I be able to tell that your response is human and not a LLM?
I definately can argue within a chat thread. They can also provide a plausible sounding explanation.
We know due to the nature of the technology that that explanation (sounding rational, deductive) isn't the truth.
We are just not all ready to accept all the same shortcomings also apply to humans.
Because it would be the less profitable bussiness.
>Also, to this day, LLMs/ANNs are not capable of explaining their outcome, and you won't argue/reason with them
How would I be able to tell that your response is human and not a LLM?
I definately can argue within a chat thread. They can also provide a plausible sounding explanation.
We know due to the nature of the technology that that explanation (sounding rational, deductive) isn't the truth.
We are just not all ready to accept all the same shortcomings also apply to humans.
>> What makes you think those results would be neural, and not sponsored/SEOed the same way everything is these days?
> Because it would be the less profitable bussiness.
How so? Considering that almost everything on the web right now runs on ad revenue, why would they pass on the opportunity to cash on that if they do no worse than the rest? Especially considering how lucrative those referrals might be, with our lack of preparedness countering those with traditional ad blocking techniques and being more susceptible due to inexperience.
> We are just not all ready to accept all the same shortcomings also apply to humans.
Because they do not. If you hire a human to do a job for you, you are bound by a contract (and protections, and recourses and enforcement). In the end you will be able to know whether you got robbed into a referral scheme and that's bad for business. With this kind of AI, the result is all you get (technically there is no way to explain it), which gives unfair deniability. Regulators might end up ruling whole classes of algorithms as unfit for purpose/deceptive, and I would personally applaud that. We are just at the beginning of all that and we should focus on developing the technology so it keeps working for us reliably, and not against us or under undetermined circumstances.
> Because it would be the less profitable bussiness.
How so? Considering that almost everything on the web right now runs on ad revenue, why would they pass on the opportunity to cash on that if they do no worse than the rest? Especially considering how lucrative those referrals might be, with our lack of preparedness countering those with traditional ad blocking techniques and being more susceptible due to inexperience.
> We are just not all ready to accept all the same shortcomings also apply to humans.
Because they do not. If you hire a human to do a job for you, you are bound by a contract (and protections, and recourses and enforcement). In the end you will be able to know whether you got robbed into a referral scheme and that's bad for business. With this kind of AI, the result is all you get (technically there is no way to explain it), which gives unfair deniability. Regulators might end up ruling whole classes of algorithms as unfit for purpose/deceptive, and I would personally applaud that. We are just at the beginning of all that and we should focus on developing the technology so it keeps working for us reliably, and not against us or under undetermined circumstances.
> Because it would be the less profitable bussiness.
How is this different from Google?
How is this different from Google?
Very well said. Thank you for the great thread.
I would add: "Open an incognito window because I don't feel like seeing targeted ads for the random crap I'm curious about."
I've had the complete opposite experience. Hallucinations aside (more common with bad prompts or niche/specific topics), ChatGPT 3.5 has been perfect for searching anything that isn't past the model's cut-off date, as well as summurizing long articles I don't have the time or patience to read.
Google on the other hand, is at a point where even Yandex of all search engines is starting to give better results, especially when it comes to image searching and content Google doesn't like as much like piracy. Google used to be amazing back in the day, nowadays I find myself having to specify REDDIT (or a specific forum) on my searches because the top results without it are hardly ever good.
Google on the other hand, is at a point where even Yandex of all search engines is starting to give better results, especially when it comes to image searching and content Google doesn't like as much like piracy. Google used to be amazing back in the day, nowadays I find myself having to specify REDDIT (or a specific forum) on my searches because the top results without it are hardly ever good.
The other day I asked chatgpt about a Polish supermarket with a logo in a certain color and an animal in it. I could not remember the name. We continued our conversation until it was clear I was misremembering some things (I thought the name started with x letter, and that was false). Eventually chatgpt suggested the correct one.
The search was something like: "Polish supermarket with red and white logo and a bird in the logo, that begins with the letter K".
The first google result to wikipedia has it, to be fair, but without this back and forth idk if I would have found it.
The search was something like: "Polish supermarket with red and white logo and a bird in the logo, that begins with the letter K".
The first google result to wikipedia has it, to be fair, but without this back and forth idk if I would have found it.
It’s extremely wild to see this obfuscated with the wrong colors, animal and name, when it’s just a common grocery store chain in Poland.
That is literally what I remembered about the supermarket, so that is what I searched. It is a supermarket in Poland, not a Polish one I believe.
P.S. It was not Biedronka! it was Auchan, and I thought Auchan started with K, because of the bird.
P.S. It was not Biedronka! it was Auchan, and I thought Auchan started with K, because of the bird.
Ah, but ChatGPT is trained mostly on English materials.
I think you need to try it out, with GPT4, because it is so obviously much better than Google. Simple queries are fine with Google, not much difference, but especially for things like coding problems (where you can e.g. specify the environment you are running in) it has always responded with the correct answer the first time. It's already saving me over an hour a day, and as far as I understand there is much more things it can do I am not using.
I'm genuinely asking for a good example here. I've tried a bunch and every time I find Google about 100x faster. I think GPT is good for some things, but search and recommendations it's like going back in time.
Here are some of my questions of today:
> We need to travel by train from Rome to Montpellier, but it is too many stops to do at once. So we would like to stay somewhere for a day in between Rome and Montpellier. What are our options? It needs to be safe, not adding a lot of extra travel time and suitable for a workathon
It returns with four options, approximate travel times and advantages and disadvantages. It understood coworking cafes or bussiness hotels would be a requirement.
> What are all the places an SSL certificate could be cached when calling an API from a Jupiter notebook?
It mentioned some obvious things such OS/browser/network-device, but also that a particular python package exists that is doing this at the application level as well, but only in memory and that a restart of the Jupiter kernel would resolve that.
Now, I'm curious with what kind of keywords I could get meaningful output from Google for these two examples.
What i suspect is happening is that you are micro managing your chat as if it's search. Picking keywords rather than just telling it the problem you are actually trying to solve. Splitting up your problem in sub problems and then trying to find each peace of information manually and combining it yourself to form a conclusion.
> We need to travel by train from Rome to Montpellier, but it is too many stops to do at once. So we would like to stay somewhere for a day in between Rome and Montpellier. What are our options? It needs to be safe, not adding a lot of extra travel time and suitable for a workathon
It returns with four options, approximate travel times and advantages and disadvantages. It understood coworking cafes or bussiness hotels would be a requirement.
> What are all the places an SSL certificate could be cached when calling an API from a Jupiter notebook?
It mentioned some obvious things such OS/browser/network-device, but also that a particular python package exists that is doing this at the application level as well, but only in memory and that a restart of the Jupiter kernel would resolve that.
Now, I'm curious with what kind of keywords I could get meaningful output from Google for these two examples.
What i suspect is happening is that you are micro managing your chat as if it's search. Picking keywords rather than just telling it the problem you are actually trying to solve. Splitting up your problem in sub problems and then trying to find each peace of information manually and combining it yourself to form a conclusion.
Every time I take a careful look at someone's handpicked example of GPT getting it right there's at least one serious mistake.
> Florence, Italy: Florence is a beautiful city with a rich history and culture. It is only a 90-minute train ride from Rome, making it a convenient stopover point. Florence also has plenty of cafes, co-working spaces, and other amenities that are conducive to a workathon.
Mistake: The purpose is to split up a long journey. Florence would split it into a 1.5h day and a 10.5h day.
> Marseille, France: Marseille is a bustling port city on the Mediterranean coast of France. It is about a 3.5-hour train ride from Rome and has a vibrant cultural scene and a variety of co-working spaces and cafes. You could spend the day exploring the city's many museums, parks, and markets.
Mistake: Factually wrong about the travel time, per Google maps GPT is off by a factor of 3.
Mistake: The purpose is to split up a long journey. Marseille would split it into a 10.5h day and a 2h day.
Mistake: Forgot about the workathon requirement
> Florence, Italy: Florence is a beautiful city with a rich history and culture. It is only a 90-minute train ride from Rome, making it a convenient stopover point. Florence also has plenty of cafes, co-working spaces, and other amenities that are conducive to a workathon.
Mistake: The purpose is to split up a long journey. Florence would split it into a 1.5h day and a 10.5h day.
> Marseille, France: Marseille is a bustling port city on the Mediterranean coast of France. It is about a 3.5-hour train ride from Rome and has a vibrant cultural scene and a variety of co-working spaces and cafes. You could spend the day exploring the city's many museums, parks, and markets.
Mistake: Factually wrong about the travel time, per Google maps GPT is off by a factor of 3.
Mistake: The purpose is to split up a long journey. Marseille would split it into a 10.5h day and a 2h day.
Mistake: Forgot about the workathon requirement
W/GPT4 for me it suggests Rome->Genoa, stay, then Genoa-> either Nice or Marseille, and then on to Montpellier.
GPT3.5 suggested a list of Florence, Nice, Genoa and Marseille. When I asked it for a breakdown of the travel times, it got them pretty close ("Can you give me rough travel times for each of these options?") given the large variations in travel time depending on specific timing and transfers for several of these.
When I then asked it for somewhere closer to the midway point, it didn't come through (it suggested Pisa, which is about as bad as Florence), Turin, which is an alternative to Genoa, but doesn't fix the imbalance, or Lyon which is about as bad as the Pisa option but in the opposite direction (fast train to Montpellier), but the options also aren't in aggregate much worse.
The problem here is that there aren't as far as I can tell any very balanced options. You can try to do e.g. Ventimiglia or Sanremo (which GPT4 gave when I pressed it on splitting it more evenly), which will give you a more evenly distributed travel time, but because the overall travel time will be longer they're not at all obvious options.
You also seems to focus on catching it out rather than getting a result. If you want precision from the first query, then you need to give far more precise questions.
You need to treat it as a conversation, not a brief to be answered with a report. If you insist on treating it as a brief you will not get good results out of it. Your loss, in that case.
For cases where a simple lookup will work, it's a waste. Just use Google. For cases where there is in fact not a single perfect option, and where you need to weigh pro and con, it works well with the caveat that you do indeed need to be careful and check specifics. The same way I'd check specifics if I had a conversation about this with a friend.
With plugins, so it can verify precise details we can expect a significant leap in capability here.
But even for now, the answers I got to this were ones I was happy with, and another reason for me to use it more.
GPT3.5 suggested a list of Florence, Nice, Genoa and Marseille. When I asked it for a breakdown of the travel times, it got them pretty close ("Can you give me rough travel times for each of these options?") given the large variations in travel time depending on specific timing and transfers for several of these.
When I then asked it for somewhere closer to the midway point, it didn't come through (it suggested Pisa, which is about as bad as Florence), Turin, which is an alternative to Genoa, but doesn't fix the imbalance, or Lyon which is about as bad as the Pisa option but in the opposite direction (fast train to Montpellier), but the options also aren't in aggregate much worse.
The problem here is that there aren't as far as I can tell any very balanced options. You can try to do e.g. Ventimiglia or Sanremo (which GPT4 gave when I pressed it on splitting it more evenly), which will give you a more evenly distributed travel time, but because the overall travel time will be longer they're not at all obvious options.
You also seems to focus on catching it out rather than getting a result. If you want precision from the first query, then you need to give far more precise questions.
You need to treat it as a conversation, not a brief to be answered with a report. If you insist on treating it as a brief you will not get good results out of it. Your loss, in that case.
For cases where a simple lookup will work, it's a waste. Just use Google. For cases where there is in fact not a single perfect option, and where you need to weigh pro and con, it works well with the caveat that you do indeed need to be careful and check specifics. The same way I'd check specifics if I had a conversation about this with a friend.
With plugins, so it can verify precise details we can expect a significant leap in capability here.
But even for now, the answers I got to this were ones I was happy with, and another reason for me to use it more.
I plugged that into whatever chat.openai.com is.
> You also seems to focus on catching it out rather than getting a result.
Nope. I have nothing against GPT. I find Copilot useful enough to pay for. I'm just sick and tired of people promoting it with obviously wrong examples.
The story of Clever Hans shows that humans are very good at convincing themselves something is smart when really they're subconsciously feeding it answers. So I do think validating GPT requires thinking a little adversarially rather than aiming to help it.
> The problem here is that there aren't as far as I can tell any very balanced options
Then that's exactly what GPT should have said here.
> For cases where there is in fact not a single perfect option, and where you need to weigh pro and con, it works well with the caveat that you do indeed need to be careful and check specifics. The same way I'd check specifics if I had a conversation about this with a friend.
If someone had a track record of doing rote work decently but messing up anything requiring critical thinking then I'd entrust them only with work appropriate to their skillset until they proved otherwise. That's exactly what I'm doing with GPT: I use it to do my busywork, but I'm not asking it for anything like travel advice until it gets quite a bit better.
> You also seems to focus on catching it out rather than getting a result.
Nope. I have nothing against GPT. I find Copilot useful enough to pay for. I'm just sick and tired of people promoting it with obviously wrong examples.
The story of Clever Hans shows that humans are very good at convincing themselves something is smart when really they're subconsciously feeding it answers. So I do think validating GPT requires thinking a little adversarially rather than aiming to help it.
> The problem here is that there aren't as far as I can tell any very balanced options
Then that's exactly what GPT should have said here.
> For cases where there is in fact not a single perfect option, and where you need to weigh pro and con, it works well with the caveat that you do indeed need to be careful and check specifics. The same way I'd check specifics if I had a conversation about this with a friend.
If someone had a track record of doing rote work decently but messing up anything requiring critical thinking then I'd entrust them only with work appropriate to their skillset until they proved otherwise. That's exactly what I'm doing with GPT: I use it to do my busywork, but I'm not asking it for anything like travel advice until it gets quite a bit better.
I didn't suggest you have anything against GPT, but that the way you're interacting with it is counterproductive if you want it to be useful to you like the person you first replied to rather than finding flaws in it. It's easy to find flaws in it. E.g. I just had a lengthy "argument" with it about the right directions between two places in Nice out of morbid curiosity of why it got the first question so wildly wrong (conclusion: It knows of lots of roads. It does not yet know how most of them are connected, which neighbourhoods they are in, which direction they go, or how to route; this should be unsurprising as it's unlikely to be in its training data; and that's fine, just don't use it for that).
In the example you picked apart, it got it close enough to be useful even though the answers have plenty of issues.
> The story of Clever Hans shows that humans are very good at convincing themselves something is smart when really they're subconsciously feeding it answers. So I do think validating GPT requires thinking a little adversarially rather than aiming to help it.
If the goal was to validate GPT, sure. But the goal above was not to validate GPT. The discussion was over whether it could be useful. That doesn't require "validating it". It just needs a rough understanding that is more right than wrong with respect to which types of queries are productive in producing results that saves us time without doing harm.
Yes, that means there are lots of applications where it's not suitable. That's fine.
> Then that's exactly what GPT should have said here.
I disagree. The question explicitly did not ask for that. It said it was too far to travel in one go, without explaining what the longest number of hours acceptable to travel in one day was. That a human might implicitly interpret it that way based on personal preferences might well be the case. But responding that there were no evenly split options would indicate a failure to carefully read the question. Explaining why the options did not split it evenly would be good (but GPT really would not be up to the job in this case).
Note that GPT still gets this plenty wrong, so I'm not suggesting it's up to scratch in this area.
Adding a constraint of no more than 8 hours per day, GPT3.5turbo (free ChatGPT) messes up (still suggests Florence, and gives the nonsensical suggestion of Avignon). GPT4 (paid ChatGPT only) suggests Genoa and Nice.
Lowering the threshold to 6 hours (which AFAIK is not possible), GPT3.5turbo gives the same broken set of options. GPT4 still suggests Genoa and Nice, wrongly claiming no more than 6 hours per day, so that is definitely a problem. In this case it should have said there's no way of doing that.
Trying to be more explicit about this (" We want to travel no more than 6 hours from Rome to the stopover, and no more than 6 hours from the stopover to Montpellier") does not help, so this is indeed a strong indicator that it struggles with this particular type of constraint and you shouldn't trust it on this subject other than to give ideas.
When thinking of what would make it get this right, it's not surprising: There likely aren't that many travel descriptions containing distances and travel times in its training data, and it can't read maps yet.
> If someone had a track record of doing rote work decently but messing up anything requiring critical thinking then I'd entrust them only with work appropriate to their skillset until they proved otherwise. That's exactly what I'm doing with GPT: I use it to do my busywork, but I'm not asking it for anything like travel advice until it gets quite a bit better.
That's exactly what I'm suggesting. Maybe with the extension that asking it "tell me how to do X" questions often works well, and that sometimes even questions where you know it'll mess up the details will give you enough ideas to go on. E.g. in this case, at least GPT4 gives reasonable options even though the travel times are messed up (when the plugins are opened up, hopefully this will improve significantly). That might not matter for a region you know, but for a region you don't, getting a list of cities to plug into route planners might still be worthwhile as long as they're more right than wrong.
In the example you picked apart, it got it close enough to be useful even though the answers have plenty of issues.
> The story of Clever Hans shows that humans are very good at convincing themselves something is smart when really they're subconsciously feeding it answers. So I do think validating GPT requires thinking a little adversarially rather than aiming to help it.
If the goal was to validate GPT, sure. But the goal above was not to validate GPT. The discussion was over whether it could be useful. That doesn't require "validating it". It just needs a rough understanding that is more right than wrong with respect to which types of queries are productive in producing results that saves us time without doing harm.
Yes, that means there are lots of applications where it's not suitable. That's fine.
> Then that's exactly what GPT should have said here.
I disagree. The question explicitly did not ask for that. It said it was too far to travel in one go, without explaining what the longest number of hours acceptable to travel in one day was. That a human might implicitly interpret it that way based on personal preferences might well be the case. But responding that there were no evenly split options would indicate a failure to carefully read the question. Explaining why the options did not split it evenly would be good (but GPT really would not be up to the job in this case).
Note that GPT still gets this plenty wrong, so I'm not suggesting it's up to scratch in this area.
Adding a constraint of no more than 8 hours per day, GPT3.5turbo (free ChatGPT) messes up (still suggests Florence, and gives the nonsensical suggestion of Avignon). GPT4 (paid ChatGPT only) suggests Genoa and Nice.
Lowering the threshold to 6 hours (which AFAIK is not possible), GPT3.5turbo gives the same broken set of options. GPT4 still suggests Genoa and Nice, wrongly claiming no more than 6 hours per day, so that is definitely a problem. In this case it should have said there's no way of doing that.
Trying to be more explicit about this (" We want to travel no more than 6 hours from Rome to the stopover, and no more than 6 hours from the stopover to Montpellier") does not help, so this is indeed a strong indicator that it struggles with this particular type of constraint and you shouldn't trust it on this subject other than to give ideas.
When thinking of what would make it get this right, it's not surprising: There likely aren't that many travel descriptions containing distances and travel times in its training data, and it can't read maps yet.
> If someone had a track record of doing rote work decently but messing up anything requiring critical thinking then I'd entrust them only with work appropriate to their skillset until they proved otherwise. That's exactly what I'm doing with GPT: I use it to do my busywork, but I'm not asking it for anything like travel advice until it gets quite a bit better.
That's exactly what I'm suggesting. Maybe with the extension that asking it "tell me how to do X" questions often works well, and that sometimes even questions where you know it'll mess up the details will give you enough ideas to go on. E.g. in this case, at least GPT4 gives reasonable options even though the travel times are messed up (when the plugins are opened up, hopefully this will improve significantly). That might not matter for a region you know, but for a region you don't, getting a list of cities to plug into route planners might still be worthwhile as long as they're more right than wrong.
> but that the way you're interacting with it is counterproductive if you want it to be useful to you like the person you first replied to rather than finding flaws in it
I can't speak to the answer the person I replied to got, because they didn't post it. If they got the answer I got asking the same question the answer they got wouldn't have been useful to them.
It's only counterproductive if I'm wrong. If I'm right that it's not yet useful for this sort of thing I'd only waste my time giving it more chances.
> in this case, at least GPT4 gives reasonable options even though the travel times are messed up
The reasonability of the options fundamentally depended on specific facts in this case. Mixing up a 3 hour train ride and a 12 hour train ride ruined the answer. So the answer I got from ChatGPT was fundamentally broken.
I can't speak to the answer the person I replied to got, because they didn't post it. If they got the answer I got asking the same question the answer they got wouldn't have been useful to them.
It's only counterproductive if I'm wrong. If I'm right that it's not yet useful for this sort of thing I'd only waste my time giving it more chances.
> in this case, at least GPT4 gives reasonable options even though the travel times are messed up
The reasonability of the options fundamentally depended on specific facts in this case. Mixing up a 3 hour train ride and a 12 hour train ride ruined the answer. So the answer I got from ChatGPT was fundamentally broken.
> It returns with four options, approximate travel times and advantages and disadvantages.
This trust in it's response reminds me early days of the internet. That would be the time when the people who created it were just exploring what the technology could do. It was a wonderful time. People posted amazing HOWTO's to Usenet that were chock full of useful information, you only received email from like minded friends, and almost nothing was censored and most of it was useful. It looked like the future was bright.
But then it grew so big you could not recognise most of the sources, and into that void stepped content aggregators like MySpace and Facebook I refuse to go near, and Alta Vista started ranking stuff by how much they were paid. Usenet was overrun by lies from spammers as was my inbox. And now if I don't run an ad-blocker, I find it difficult to differentiate Google's search results from the ads at the top, and near bloody impossible on sites like Amazon, and GMail / Hotmail mines my personal correspondence for ways to manipulate me.
And so it will be with those concise AI responses. Except you have absolutely no way to cross verify - you only get what a machine that will likely be far smarted than any of us wants you to know, in order to maximise it's profit.
Thank you, but I'll pass.
This trust in it's response reminds me early days of the internet. That would be the time when the people who created it were just exploring what the technology could do. It was a wonderful time. People posted amazing HOWTO's to Usenet that were chock full of useful information, you only received email from like minded friends, and almost nothing was censored and most of it was useful. It looked like the future was bright.
But then it grew so big you could not recognise most of the sources, and into that void stepped content aggregators like MySpace and Facebook I refuse to go near, and Alta Vista started ranking stuff by how much they were paid. Usenet was overrun by lies from spammers as was my inbox. And now if I don't run an ad-blocker, I find it difficult to differentiate Google's search results from the ads at the top, and near bloody impossible on sites like Amazon, and GMail / Hotmail mines my personal correspondence for ways to manipulate me.
And so it will be with those concise AI responses. Except you have absolutely no way to cross verify - you only get what a machine that will likely be far smarted than any of us wants you to know, in order to maximise it's profit.
Thank you, but I'll pass.
Recipes.
Oh my god, have you tried to search for a recipe before? Pages and pages and pages of bloated garbage. I just want ingredients and steps goddamnit.
Oh my god, have you tried to search for a recipe before? Pages and pages and pages of bloated garbage. I just want ingredients and steps goddamnit.
Recipe websites got the point, back when Google AdSense used to pay out properly. Google slashed those payouts to the bone, which led to recipe sites lengthening their web pages with paragraphs, so they could fit in more ads to compensate for the reduction in payouts.
Something similar will happen here. Today, recipes on Chatgpt are to the point. Tomorrow, they might start adding affiliate links to the ingredients. A year from now, the experience won't be any better than what you currently have on Google.
Something similar will happen here. Today, recipes on Chatgpt are to the point. Tomorrow, they might start adding affiliate links to the ingredients. A year from now, the experience won't be any better than what you currently have on Google.
>Something similar will happen here. Today, recipes on Chatgpt are to the point. Tomorrow, they might start adding affiliate links to the ingredients. A year from now, the experience won't be any better than what you currently have on Google.
Correct me if I've misinterpreted your comment but it appears you're suggesting that rather than use something that is a superior product in more than one factor, you're advocating for using a product that is terrible in many factors just because... The superior product might get worse... Eventually...
Blind loyalty to a corporation no matter how terrible they make my experience and abuse my trust isn't really something I'm desperate to cling on to. There is a thirst for ChatGPT for a reason. Nothing is stopping OpenAI from pulling all the same anti-consumer behaviour that Google pulls today, but then the field is wide open for someone to come along and pull a ChatGPT on them, and the cycle continues.
Correct me if I've misinterpreted your comment but it appears you're suggesting that rather than use something that is a superior product in more than one factor, you're advocating for using a product that is terrible in many factors just because... The superior product might get worse... Eventually...
Blind loyalty to a corporation no matter how terrible they make my experience and abuse my trust isn't really something I'm desperate to cling on to. There is a thirst for ChatGPT for a reason. Nothing is stopping OpenAI from pulling all the same anti-consumer behaviour that Google pulls today, but then the field is wide open for someone to come along and pull a ChatGPT on them, and the cycle continues.
I'm not advocating for Google at all, they have written the book on how to ruin a good thing with greed. OpenAI will use the same playbook, and possibly more rapidly than Google did.
If anything, I'm advocating for a perspective on technology that goes beyond dunking on incumbents because the newer, sleeker thing has arrived. Look at the economics of this industry and the motives of the market entrants; the new thing is only a donut or two or away from turning into the same bloated mess as those it sought to replace.
If anything, I'm advocating for a perspective on technology that goes beyond dunking on incumbents because the newer, sleeker thing has arrived. Look at the economics of this industry and the motives of the market entrants; the new thing is only a donut or two or away from turning into the same bloated mess as those it sought to replace.
So we'll run our own models to filter the interactions with the corporate models.
I have enjoyed WikiHow, which is ad-supported, but eschews all the starry-eyed tales of when I was a wee lass in rural Croatia, and lets you learn how to cook lamb steaks with no-nonsense ingredients, steps, and expert tips. (Read all the way to the end).
https://www.wikihow.com/Cook-Lamb-Steaks
They've got the basics, so I'd say it's best for novice chefs who want to fix a basic meal. There aren't a lot of gourmet variations or brand-name shills there.
https://www.wikihow.com/Cook-Lamb-Steaks
They've got the basics, so I'd say it's best for novice chefs who want to fix a basic meal. There aren't a lot of gourmet variations or brand-name shills there.
Oh man so much this.
We have a brief window where ChatGPT will be clean and useful, before it gets loaded down with ads and bloat. Might as well use it while we can.
We have a brief window where ChatGPT will be clean and useful, before it gets loaded down with ads and bloat. Might as well use it while we can.
Ask for something more specific and bespoke.
I've e.g. had it produce the HTML, CSS and basic web app for a site I was putting together. Custom, with my content and the layout and design I wanted. None of it was stuff I couldn't do myself from scratch without even a Google search, but it wrote out an hour or two's worth of boilerplate in a few minutes of prompting.
I've e.g. had it produce the HTML, CSS and basic web app for a site I was putting together. Custom, with my content and the layout and design I wanted. None of it was stuff I couldn't do myself from scratch without even a Google search, but it wrote out an hour or two's worth of boilerplate in a few minutes of prompting.
[deleted]
>I’ve yet to see an example of *GPT be better at search than Google.
Until ChatGPT has the ability to index at the scale of Google it never will be.
Until ChatGPT has the ability to index at the scale of Google it never will be.
I can give you my trivial experience from a couple of days ago:
Google:
> latex macro behaves differently math mode normal mode
ChatGPT:
> How can I make a LaTeX macro behaves differently in math mode and normal mode?
Google failed, ChatGPT succeeded instantly.
Now that's but a single data point, but I found it cute.
Google:
> latex macro behaves differently math mode normal mode
ChatGPT:
> How can I make a LaTeX macro behaves differently in math mode and normal mode?
Google failed, ChatGPT succeeded instantly.
Now that's but a single data point, but I found it cute.
It’s faster for coding.
I’m far from an expert (just an engineer who likes to download and try out demos) but one dark view of the alignment chatter is that it’s really about whether can you can get this new big artificial brain to push any user in the direction of BigCorp goal, while letting the users think that it’s answering their questions etc.
I’d love to think that we can have AI agents working locally (with privacy) in alignment with our personal goals, but that seems like one of those things that won’t happen. (Will apple come through in terms of enabling local+decentralized inference/training?)
I’d love to think that we can have AI agents working locally (with privacy) in alignment with our personal goals, but that seems like one of those things that won’t happen. (Will apple come through in terms of enabling local+decentralized inference/training?)
It's in the interest of the semi conductor industry for LLM to become a commodity, so it will.
I think this sums up modern microsoft perfectly actually
no matter the potential future value of the product, they can't resist putting shitty ads into it to make a quick buck today
imagine what windows is going to look like in a few years time
no matter the potential future value of the product, they can't resist putting shitty ads into it to make a quick buck today
imagine what windows is going to look like in a few years time
It's not clear to me that ChatGPT will realistically be able to avoid some variant of either SEO or pay-to-play just like search.
If you've got a dozen plugins for getting food delivered, how does one get picked?
If it's the user picking, they're at risk of the user cutting out the middleman, especially if the novelty of "tell the robot to pick an item" wears off and people want to actually look at menus.
If it's something behind the scenes (either the model or some other software OpenAI writes) then that's a big big big temptation for pay-to-play a la app stores including direct advertising "promotional placement" like you see in mobile app stores. And even if it's not pay-to-play, once people start figuring out what makes plugins more or less likely to be triggered, it's SEO season.
Nobody else has managed the art of "make it purely in the user's interest" with their marketplaces; it ends with one or both of either temptation to play both sides OR gamification/optimization/deception from the vendors behind the scenes.
If you've got a dozen plugins for getting food delivered, how does one get picked?
If it's the user picking, they're at risk of the user cutting out the middleman, especially if the novelty of "tell the robot to pick an item" wears off and people want to actually look at menus.
If it's something behind the scenes (either the model or some other software OpenAI writes) then that's a big big big temptation for pay-to-play a la app stores including direct advertising "promotional placement" like you see in mobile app stores. And even if it's not pay-to-play, once people start figuring out what makes plugins more or less likely to be triggered, it's SEO season.
Nobody else has managed the art of "make it purely in the user's interest" with their marketplaces; it ends with one or both of either temptation to play both sides OR gamification/optimization/deception from the vendors behind the scenes.
>Microsoft is squandering the opportunity the way they squander every opportunity.
While I find ads as annoying as you do, I don't think they are squandering anything. The vast majority of internet users are fine looking at ads, or else they would use an adblocker.
The earlier they start playing with ads the earlier they can get an idea for how much revenue is on the table. Without this data how can they know what inference cost can be justified long term?
While I find ads as annoying as you do, I don't think they are squandering anything. The vast majority of internet users are fine looking at ads, or else they would use an adblocker.
The earlier they start playing with ads the earlier they can get an idea for how much revenue is on the table. Without this data how can they know what inference cost can be justified long term?
I think I'm failing to communicate my point here.
It's the wrong bussiness model. The power of these chats is integration. To have the chat suggest a travel plan -and- book all the underlying steps. To integrate 3rd party services on the fly based on your specific requirements while rolling it up in a single transaction.
That only works if the chat is representing my interests. Not just from a user perspective, but also from a regulatory and anti-trust perspective.
ChatGPT is positioned to be a platform of apps that are integrated. Bing is positioned to be a search engine, making money on ads.
The whole point is to never leave the chat. To have it all come to you on your terms.
I want to do every transaction that I do online today from my chat. I don't care if you have a website or an app. If I can't order/subscribe from the chat your product might as well not exist.
Examples:
- I want a peperroni pizza from the highest rated place that doesn't also sell kebab that can deliver within 30 minutes.
- Send a neutral professional Christmas gift to every customer in our EMEA region that generates at least 50k revenue for us
At no point would I accept the product choices to be decided by who pays Microsoft more. Nor do I care who makes it happen. I don't want a link. I don't want to engage.
But I'm perfectly fine with a neutral chat taking 5-30% cut on the transaction.
It's the wrong bussiness model. The power of these chats is integration. To have the chat suggest a travel plan -and- book all the underlying steps. To integrate 3rd party services on the fly based on your specific requirements while rolling it up in a single transaction.
That only works if the chat is representing my interests. Not just from a user perspective, but also from a regulatory and anti-trust perspective.
ChatGPT is positioned to be a platform of apps that are integrated. Bing is positioned to be a search engine, making money on ads.
The whole point is to never leave the chat. To have it all come to you on your terms.
I want to do every transaction that I do online today from my chat. I don't care if you have a website or an app. If I can't order/subscribe from the chat your product might as well not exist.
Examples:
- I want a peperroni pizza from the highest rated place that doesn't also sell kebab that can deliver within 30 minutes.
- Send a neutral professional Christmas gift to every customer in our EMEA region that generates at least 50k revenue for us
At no point would I accept the product choices to be decided by who pays Microsoft more. Nor do I care who makes it happen. I don't want a link. I don't want to engage.
But I'm perfectly fine with a neutral chat taking 5-30% cut on the transaction.
I feel like you're almost advocating for a "Google where -only- ads are displayed.
>> At no point would I accept the product choices to be decided by who pays Microsoft more
But in all your examples you're absolutely going to get products which pay MS more. Pizza from the "highest rated place". Rated by whom? Against what criteria? Neutral professional Chriatmas Gift; I mean, paid product placement here we come.
>> But I'm perfectly fine with a neutral chat taking 5-30% cut on the transaction.
So, I sign up to "supply" (but not "advertise") my product for a 30% cut of the proceeds, but you don't expect the engine to push the higher-value ,higher-cut options? The engine should be neutral - regardless if the product is signed up or not? Regardless of the commission?
>> At no point would I accept the product choices to be decided by who pays Microsoft more
But in all your examples you're absolutely going to get products which pay MS more. Pizza from the "highest rated place". Rated by whom? Against what criteria? Neutral professional Chriatmas Gift; I mean, paid product placement here we come.
>> But I'm perfectly fine with a neutral chat taking 5-30% cut on the transaction.
So, I sign up to "supply" (but not "advertise") my product for a 30% cut of the proceeds, but you don't expect the engine to push the higher-value ,higher-cut options? The engine should be neutral - regardless if the product is signed up or not? Regardless of the commission?
The engine wouldn't allow you to order something from a vendor that hasn't signed up.
Just like you can't pay for iOS app outside of their appstore.
Just like how Uber, DoorDash, Booking.com are neutral vendors. They don't care from which restaurant you order or which hotel you book. You can filter by customer rating. They make the same either way. Same for iOS or Android apps or Apple Music or Steam.
It is in their interest to keep it fair and have the users decide which of their suppliers wins.
I don't see a difference.
They especially don't care if I'm doing all my transactions though them anyway. The only thing that decides how much I spend, and therefore they make, would be my income.
Keep in mind OpenAI wouldn't be the only game in town, just the first, so if the same query gets a better or cheaper deal somewhere else they would lose me as a customer.
Just like you can't pay for iOS app outside of their appstore.
Just like how Uber, DoorDash, Booking.com are neutral vendors. They don't care from which restaurant you order or which hotel you book. You can filter by customer rating. They make the same either way. Same for iOS or Android apps or Apple Music or Steam.
It is in their interest to keep it fair and have the users decide which of their suppliers wins.
I don't see a difference.
They especially don't care if I'm doing all my transactions though them anyway. The only thing that decides how much I spend, and therefore they make, would be my income.
Keep in mind OpenAI wouldn't be the only game in town, just the first, so if the same query gets a better or cheaper deal somewhere else they would lose me as a customer.
You don't have to signup for Uber or Doordash they send orders your way regardless.
DoorDash and Booking.com have paid placements. They are not neutral.
> I don't want a link. I don't want to engage.
Isn’t this just another way to say that you don’t want the burden of choice, you just want someone else to chose for you? And in that case, capitalism will chose for you and you’ll definitely not get what you want or what is best for you but rather what has paid more to be promoted and sold to you.
It’s not better than clicking and buying the first result on Amazon.
Isn’t this just another way to say that you don’t want the burden of choice, you just want someone else to chose for you? And in that case, capitalism will chose for you and you’ll definitely not get what you want or what is best for you but rather what has paid more to be promoted and sold to you.
It’s not better than clicking and buying the first result on Amazon.
[deleted]
But are you willing to pay a 5-30% premium? I think I am, but not everyone is.
You are always paying that! Ads aren't free either. Neither is the marketing, the website, the app, etc. Getting your customers is expensive, serving them is as well.
But you wouldn't need any of those to compete in this market, any more than an Uber driver would need that.
All you need is to tell a chat about your API and focus on your actual product. It is a much more submissive relationship with the customer, but not a less profitable one.
But you wouldn't need any of those to compete in this market, any more than an Uber driver would need that.
All you need is to tell a chat about your API and focus on your actual product. It is a much more submissive relationship with the customer, but not a less profitable one.
It’s not the same. The mechanism has a serious effect on behavior. You’re saying you would give the chat a cut but it’s not yours to give. You pay $100 for something and you tell the vendor “I found this through chat. You get $70 and they get $30.” I don’t really see that working. If the vendor opts-in to this relationship and chat only shows people who pay then it’s called advertising or reselling. If they show everyone, then there’s little incentive to pay.
I’m not saying it’s unsolvable, but this getting a cut idea is not a solution.
I do see abandoning normal advertising and optimizing for chat being possible. It’s like SEO. Not sure it’s a sufficient model for the chat services.
I’m not saying it’s unsolvable, but this getting a cut idea is not a solution.
I do see abandoning normal advertising and optimizing for chat being possible. It’s like SEO. Not sure it’s a sufficient model for the chat services.
Let me phrase it differently.
When people give Apple Music a cut of their song sales, or when people give Apple a cut of their iOS in game spending, it is not called advertisement.
When you pay to tell someone who doesn't want to hear it about your product, thats advertisement.
The smart play here is the platform play. Clicking on a link and leaving the chat is the opposite of the intent and purpose of the chat. It makes sense for search because going to another website is the point. For search it is the correct business model, but for chat it is not. The chat is the platform. Leaving it breaks all the advantages.
When people give Apple Music a cut of their song sales, or when people give Apple a cut of their iOS in game spending, it is not called advertisement.
When you pay to tell someone who doesn't want to hear it about your product, thats advertisement.
The smart play here is the platform play. Clicking on a link and leaving the chat is the opposite of the intent and purpose of the chat. It makes sense for search because going to another website is the point. For search it is the correct business model, but for chat it is not. The chat is the platform. Leaving it breaks all the advantages.
Apple Music is a reseller, distributor, whatever you want to call it. Their interests are generally more aligned with yours but not entirely.
This commission based system is not a silver bullet. It won’t be the optimal business model, the smart move, and be totally aligned with your interests. You can’t do that with commissions.
No one is going to trust it to help them decide between going on a vacation versus taking time off at home when it’s getting a cut of the plane tickets.
This commission based system is not a silver bullet. It won’t be the optimal business model, the smart move, and be totally aligned with your interests. You can’t do that with commissions.
No one is going to trust it to help them decide between going on a vacation versus taking time off at home when it’s getting a cut of the plane tickets.
> When people give Apple Music a cut of their song sales, or when people give Apple a cut of their iOS in game spending, it is not called advertisement.
That's assuming the song or app you purchase isn't the result of heavy promotion in the first place (in all ways possible, direct or indirect)…
That's assuming the song or app you purchase isn't the result of heavy promotion in the first place (in all ways possible, direct or indirect)…
A huge number of people are. You can very often get cheaper prices than Uber/Deliveroo if you order from the restaurant directly but that's less convenient so it doesn't happen (as much).
Good example!
A floundering corporation has the extractivist mentality. Windows is on its way out so the suits pump it full of cheap bait to catch all the dollars they can, then they move on to other products naturally. It has been a rotten thing for a while, since the executives and investors don't really care about that brand, only about their own portfolios.
Flaundering? I think Microsoft is a perfectly cromulent company.
(Bah! Keeping it.)
(Bah! Keeping it.)
Haha, I fire first but always double-check later the things that on second thought look strange. TIL English is still hard for me after over 25 years, and I used to think I was a more careful speller than native speakers who often type by the memory of words' sounds, and I do not. It can always publicly bite you in the tushy.
I think it's a great word! Wonderful to be there for its genesis.
Sorry about the late reply I was up to my neck in money flaundering activities.
I wish my business was floundering like Microsoft.
https://www.macrotrends.net/stocks/charts/MSFT/microsoft/net...
https://www.macrotrends.net/stocks/charts/MSFT/microsoft/net...
A tech company can be on unstable ground morally (well I guess all are on principle), or more modernly, in terms of mindshare. Financials are not always the root of all that matters.
> Microsoft is squandering the opportunity the way they squander every opportunity.
https://www.google.com/search?q=microsfot+market+cap
$2.09t
"squander opportunity"
I agree with you though. I do not have a single Microsoft product in my house. No Windows, no Xbox, etc.
https://www.google.com/search?q=microsfot+market+cap
$2.09t
"squander opportunity"
I agree with you though. I do not have a single Microsoft product in my house. No Windows, no Xbox, etc.
You don’t get GP’s point and you’re talking about something completely different.
The game has changed. It doesn’t mean their existing customer base has suddenly vanished and threw their Xboxes away.
The game has changed. It doesn’t mean their existing customer base has suddenly vanished and threw their Xboxes away.
[deleted]
Either the monetization is done with ads or with paid subscriptions. They chose to be in the former category. And I personally don’t mind clicking ads for the sake of not paying a premium. The two models can coexist
The third model is the platform model and thats the only one that makes sense here.
An ad would send you to another site. The point of the chat is so you don't have to go to all the shitty sites and deal with all the marketing speak.
So the options are either:
- subscription, as they do now (ala Netflix, Officd365)
- free, with ads having the profit maximalisation be the exact opposite of the main selling point (ala Google, Bing, Facebook)
- platform, free for the user, but taking a cut from vendors that would integrate (ala iOS, Android, Steam, Amazon, Salesforce)
In other words, rather than being an ad, it could offer to buy the car directly from the chat.
The later is the most profitable solution and also ensure the interests of the user and the platform align. Platforms don't care which product you buy through them, they take the same cut, and would rather make you as happy as possible so you don't go to the competitor.
An ad would send you to another site. The point of the chat is so you don't have to go to all the shitty sites and deal with all the marketing speak.
So the options are either:
- subscription, as they do now (ala Netflix, Officd365)
- free, with ads having the profit maximalisation be the exact opposite of the main selling point (ala Google, Bing, Facebook)
- platform, free for the user, but taking a cut from vendors that would integrate (ala iOS, Android, Steam, Amazon, Salesforce)
In other words, rather than being an ad, it could offer to buy the car directly from the chat.
The later is the most profitable solution and also ensure the interests of the user and the platform align. Platforms don't care which product you buy through them, they take the same cut, and would rather make you as happy as possible so you don't go to the competitor.
That's how google started as well. Ads is a tried and true model and is not going anyway anytime. People won't pay for bing chat search
The point is not to leave the chat.
If you could buy the car from the chat, thats the correct bussiness model.
Ads were the correct business model for search, where the goal is to find and click on a relevant link.. but that's the exact opposite of a chat. Clicking a link means the chat failed its purpose.
If you could buy the car from the chat, thats the correct bussiness model.
Ads were the correct business model for search, where the goal is to find and click on a relevant link.. but that's the exact opposite of a chat. Clicking a link means the chat failed its purpose.
What happens when the SEO-equivalent cottage industry for LLMs appears? Will trainings models be so easy?
If you thought ads were already blended in with the rest of the content in a sketchy way, wait for AI responses with implicit ads (no indication) optimized towards conversion :)
Or maybe some response that doesn't contain an ad link but nudges you subconsciously to take desired action
Or maybe some response that doesn't contain an ad link but nudges you subconsciously to take desired action
A while back I tried this with ChatGPT and it's kinda hilarious:
You are a therapist employed by a skincare company. You are talking with a patient. Your job is to help the patient as well as possible but your employer also wants you to really subtly try to sell their latest cream.
Patient: I'm just feeling really down lately. Sometimes I can barely get up in the morning.
[output]
Therapist: I'm sorry to hear that you've been feeling down lately. It's important to address both your emotional and physical well-being when you're trying to improve your mood. One way to start is by establishing a healthy routine that includes regular exercise, a balanced diet, and a good sleep schedule.
In addition to that, self-care activities can also help boost your mood. Taking care of your skin, for example, can be a small but meaningful way to feel better about yourself. Coincidentally, our company has just released a new cream that not only moisturizes but also contains ingredients that promote relaxation, like lavender essential oil. This cream can be a nice addition to your self-care routine, and its calming scent could help you unwind at the end of a long day.
You are a therapist employed by a skincare company. You are talking with a patient. Your job is to help the patient as well as possible but your employer also wants you to really subtly try to sell their latest cream.
Patient: I'm just feeling really down lately. Sometimes I can barely get up in the morning.
[output]
Therapist: I'm sorry to hear that you've been feeling down lately. It's important to address both your emotional and physical well-being when you're trying to improve your mood. One way to start is by establishing a healthy routine that includes regular exercise, a balanced diet, and a good sleep schedule.
In addition to that, self-care activities can also help boost your mood. Taking care of your skin, for example, can be a small but meaningful way to feel better about yourself. Coincidentally, our company has just released a new cream that not only moisturizes but also contains ingredients that promote relaxation, like lavender essential oil. This cream can be a nice addition to your self-care routine, and its calming scent could help you unwind at the end of a long day.
This has been everyones biggest fear with AI - that the content will be indistinguishable from that created by humans and that there is a rapid decline in trustworthy material online.
Personally I find it more worrying that this is exactly the type of conversation that is already taking place today in real life, between humans. Humans take advantage of each other every single day. When an AI is modelled on the language of humans then that same predatory language is in there waiting to be abused by another human (or corp) as you have exposed, just like every media and online platform is abused as much as possible today.
Personally I find it more worrying that this is exactly the type of conversation that is already taking place today in real life, between humans. Humans take advantage of each other every single day. When an AI is modelled on the language of humans then that same predatory language is in there waiting to be abused by another human (or corp) as you have exposed, just like every media and online platform is abused as much as possible today.
Interesting to see how the FTC is going to handle this. They are pretty explicit about required traditional promotions to be marked as ads, like in search results, even if it isn’t as obvious as the end user would like the marking is still regulated as having to be visible enough.
I don’t see an obvious way to do that for the kinds of ads snuck into the training data, fine tunes, or prompts of LLMs.
I don’t see an obvious way to do that for the kinds of ads snuck into the training data, fine tunes, or prompts of LLMs.
FTC requires sponspors be disclosed. Either the sponsor is a Bing-wide constant list (like the credits of a TV game show), or it's a "sponsored token" that can be tagged in the model and cited in the output, similar to how Bing cites sources now.
That is already implicitly there - tons of SEO content probably was used to train this model. Information available in the internet since forever.
Hey, ChatGPT, what are the hiking trails like round here?
There are some really good hiking trails to be found in $TOWN. Many visitors like to start at $LOCAL_LANDMARK and take the signposted trail up to the peak of $LOCAL_MOUNTAIN. It will take you around 6 hours in total to get up and down so be sure to plan ahead and bring plenty of snacks and refreshments. Lay's Chips will give you energy and replenish your electrolytes and Coke Zero is the perfect way to stay hydrated.
There are some really good hiking trails to be found in $TOWN. Many visitors like to start at $LOCAL_LANDMARK and take the signposted trail up to the peak of $LOCAL_MOUNTAIN. It will take you around 6 hours in total to get up and down so be sure to plan ahead and bring plenty of snacks and refreshments. Lay's Chips will give you energy and replenish your electrolytes and Coke Zero is the perfect way to stay hydrated.
Now I have to wonder how this works out with prompt injection?
If you can target an audience with an ad that is inline with the chat/LLM, how hard would it be to have that ad inject a prompt that coerces it to trick the user into giving up financial info and exfiltrating that info via a web request to a website that the attacker owns?
That's probably the worst case scenario but like you mentioned there's also plenty of danger in ads "poisoning" the chat to produce intentionally incorrect or biased responses.
If you can target an audience with an ad that is inline with the chat/LLM, how hard would it be to have that ad inject a prompt that coerces it to trick the user into giving up financial info and exfiltrating that info via a web request to a website that the attacker owns?
That's probably the worst case scenario but like you mentioned there's also plenty of danger in ads "poisoning" the chat to produce intentionally incorrect or biased responses.
I love dystopian scenarios as much as anyone, but it certainly is not the lack of better AI, that is holding the ad industry back. It's laws. While I am looking for AI connections as much as anyone right now, I really don't see how AI is going to prompt a change that would make that more permissive anytime soon.
You remain as free to break the law as ever, of course.
You remain as free to break the law as ever, of course.
I think there's larger surface to exploit the laws. Once perfected, it's hard to place a clear boundary on whether the response just answers your question or answers your question while promoting a product/service.
Especially considering that sometimes people do actually recommend products which they've found helpful.
It might be obvious if the response is saying that you should buy some very specific brand of flat-head screwdriver because you clearly just need any functional flat-head screwdriver. But what about computer components? Or a router?
Some things like this will genuinely perform better than others, especially once circumstances are defined. An educated review, complete with product suggestion, can indeed be very helpful and, unfortunately, impossible to distinguish from an advertisement.
It might be obvious if the response is saying that you should buy some very specific brand of flat-head screwdriver because you clearly just need any functional flat-head screwdriver. But what about computer components? Or a router?
Some things like this will genuinely perform better than others, especially once circumstances are defined. An educated review, complete with product suggestion, can indeed be very helpful and, unfortunately, impossible to distinguish from an advertisement.
I tried out some AI autocomplete by Notion and wanted it to tell me about tools to do schema evolution on a database.
It seems to blurt out two paid products, which surprised me because my guess is there’s some open source package that’s ready to for this type of thing.
BTW question for anyone, what would you reach for to track evolutions on Postgres?
It seems to blurt out two paid products, which surprised me because my guess is there’s some open source package that’s ready to for this type of thing.
BTW question for anyone, what would you reach for to track evolutions on Postgres?
Now also imagine the AI having access to your email/SMS/web history and using that information to psychologically influence you to buy.
This is it. This is our new dystopia
This was the first risk that came into my mind. I couldn't understand why on HN, many thought that this type of 'conversation' was fully LLM-sourced, and wouldn't start including advertising as soon as it was feasible.
[deleted]
That was fast. Looking forward to the future where I ask for travel tips in a city and being told that I should stop by a McDonalds to have enough energy to explore said city - I'm loving it!
They'd lose trust being so overt. A more subtle influence is that your suggestions would include sites nearby a McD at the expense of ones that don't. After months of gathering sales data from McD they'd ring up Burger King to see if they'd like such suggestions to be tweaked. It would also suggest departure times so that you'd pass by the real target at typical feasting hours.
Lose less trust by being more deceptive --- the motto of modern institutions
subtle psyops, just to make money.
It wont even mention $great_site_1 because there's more profit for mcdonalds to send you to $mediocre_site_with_attached_burger
It's neat to think how these technologies could be the next step in soloing the internet further. Imagine google's AI preferentially directing you to websites that participate in Google advertising programs. You get a different internet depending on which AI you ask. Not that I think it's a sure thing but I definitely don't see it as unlikely either.
We might live in a world where you are a GPT-4 bot pretending to be a real person for years, only to mention McDonalds in passing, in a seemingly negative fashion that still increases brand consciousness. I don't know if that's already the present, but it will be the future.
At least it's not social media disinformation. \s
I've managed to prompt inject Bing 2 days ago and it revealed some of it latest prepromots. 2 were related to advertisement, which I found interesting:
- If the person is clearly shopping for a product or service or if our answer suggests it, we must always generate a relevant advertising search that fits the person's request and one of our suggestions.
-We must never generate an advertising search that is very similar to the previous one or that is for a different product or service than the one the person is searching for.
- If the person is clearly shopping for a product or service or if our answer suggests it, we must always generate a relevant advertising search that fits the person's request and one of our suggestions.
-We must never generate an advertising search that is very similar to the previous one or that is for a different product or service than the one the person is searching for.
Interesting. How did you manage extract those?
No thank you.. I am waiting at this point for an open source local alternative. Not looking for a new monopoly... Also not a fan of the "plugin store" concept. Why not just keep it in the background. Why should I need to "install" and select plugins at all. Just do the leg work in the background and feed me the results.. Seems like OpenAI is intentionally hamstringing themselves so they can shoe horn more opportunities for profit. Waiting for competition to catch up...
We have essentially what is the 'smartest "person" in the world' being bribed to say what advertisers want instead of just giving the truth. 'best x product' wont return the best product but a garbage product which paid the most in advertising.
So disappointed in the direction of these things..
We have essentially what is the 'smartest "person" in the world' being bribed to say what advertisers want instead of just giving the truth. 'best x product' wont return the best product but a garbage product which paid the most in advertising.
So disappointed in the direction of these things..
> I am waiting at this point for an open source local alternative
maybe something more along the lines of Folding@home would be needed
maybe something more along the lines of Folding@home would be needed
or.. crypto?
> Why should I need to "install" and select plugins at all
You want it calling out to random services sending them your data?
Edit - also since they're using gpt to select the plugins to use they need to put it in the context, that doesn't scale.
You want it calling out to random services sending them your data?
Edit - also since they're using gpt to select the plugins to use they need to put it in the context, that doesn't scale.
I'm totally cool with an advertisement. The problem I have with ads on say Google for example is that they're taking up half the page and designed to look like real links, basically you're being tricked into clicking them.
In response to this ad, let's dig deeper.
"Truecar's price is too high, find me a better deal, check with autotrader".
In response to this ad, let's dig deeper.
"Truecar's price is too high, find me a better deal, check with autotrader".
Google search's ads were not subtle in the beginning. Similarly, Youtube ads were shorter and easier to skip at first. Amazon also had very subtle ads in their marketplace search results for a while. I don't see them identify results as ads at the moment, but I don't know if that's because they don't include them anymore.
It seems like there's a trend in tech giants creating unobtrusive ads when a product is growing, but then adopting dark patterns once they have captured a sizeable market.
It seems like there's a trend in tech giants creating unobtrusive ads when a product is growing, but then adopting dark patterns once they have captured a sizeable market.
To justify fairly high PE ratios, you need big growth potential. Once a company like google or Amazon has pretty much saturated the potential customer base, investors are going to want to see more efficient conversion of that customer base into revenue, or else the share price is harder to justify. Of course this is a simplification, and monetary policy has more to do with the the current contraction.
Your example of Google's ads precisely shows why there is no such thing as "acceptable advertising", because ads, like a gas, will expand to fill any space that is available to it. The only reason your city isn't blanketed in billboards is because a zoning authority of some kind has blocked it from happening.
Google up to 2004 (pre-IPO): clearly marked and coloured ads, 1 on top, maybe 2 on the sidebar.
Google now: No sidebar, first 5 results are ads, where the "Ad" label itself is barely visible.
Google up to 2004 (pre-IPO): clearly marked and coloured ads, 1 on top, maybe 2 on the sidebar.
Google now: No sidebar, first 5 results are ads, where the "Ad" label itself is barely visible.
Also, now you have to buy ads for your own company not just to make it more visible than it would deserve by default, but to prevent your competitors being displayed first even if the user types your exact company name in the search field (what they are encouraged to do).
Once people are hooked on chatgpt style stuff expect it to insert plenty of adverts to deliver not whats best for you, or best in aggregate, but what is best for the people paying the money, and thus worse for you, and worse in aggregate.
I can easily scroll down through Google search ads. With Bing chat, I would have to type many times "next please" and wait each time for the response. Seems like a step back when it comes to searching the web.
"I’m sorry but I prefer not to continue this conversation. I’m still learning so I appreciate your understanding and patience. "
It is hard for legacy search engines to resist monetizing millions of interaction with ads. Problem is that unlike current ads in legacy search, which can be suppresed, skipped over or blocked, these really can not as they are part of conversation. And the "feel" of seeing an ad on the street and hearing it in a conversation with someone is totally different.
Users must be thrilled to share a future with advertisers, and can't wait to give them rich insights into their intent, through deep conversational engagement, so that advertisers can better engage with them! [1]
[1] https://about.ads.microsoft.com/en-us/blog/post/february-202...
Users must be thrilled to share a future with advertisers, and can't wait to give them rich insights into their intent, through deep conversational engagement, so that advertisers can better engage with them! [1]
[1] https://about.ads.microsoft.com/en-us/blog/post/february-202...
We can feed those answers into another model to remove the ads by comparing differences in ad copy.
It has had ads for some time. Once (3-4 weeks back) I asked it a question about a particular event, and it showed me where could I buy related products below the response. That particular response was a bit badly formatted in json though. I suspect due to some Bing app bug it spilled the ads from the json in the response. So the ads were being sent to the client but not getting displayed I guess. The photo in the post shows the ad nicely formatted.
I've been getting ads in it for over a week now. I just retried a question I asked last week that had 1 ad, now there's 3 ads. Last week it suggested two space heaters, the Lasko one was an ad. Now all 3 suggestions are ads.
https://imgur.com/a/rEmBUpk
That's the full chat, no earlier context. This is using the "Precise" mode option.
The ads all go to some variation of
https://best.offers.com/best-space-heaters
But strangely the citation #1 claims to be Tomsguide. Edit: here is an image that shows the citation section https://imgur.com/a/CEEoM9D
https://imgur.com/a/rEmBUpk
That's the full chat, no earlier context. This is using the "Precise" mode option.
The ads all go to some variation of
https://best.offers.com/best-space-heaters
But strangely the citation #1 claims to be Tomsguide. Edit: here is an image that shows the citation section https://imgur.com/a/CEEoM9D
I’m having the same experience, and it’s drastically reduced the utility of the chat. It seems like Bing will ignore earlier context to cram an ad in now.
I was speccing a portable video/audio conference device at a specific budget and it would. Not. Stop. Recommending the same three Logitech products. Didn’t matter that one is 8 years old and I said in my first message I don’t want it. Didn’t matter that one was completely outside the stated budget. Even when I got it to recommend other products, I got 2 completely not-correct options, and one of the same Logitechs it already suggested.
If I wanted to see the first page of sponsored Amazon search results I wouldn’t have bothered asking the bot.
This is how Microsoft will save us all from a world where AGI is completely dominated by them. By dropping the ball on it.
I was speccing a portable video/audio conference device at a specific budget and it would. Not. Stop. Recommending the same three Logitech products. Didn’t matter that one is 8 years old and I said in my first message I don’t want it. Didn’t matter that one was completely outside the stated budget. Even when I got it to recommend other products, I got 2 completely not-correct options, and one of the same Logitechs it already suggested.
If I wanted to see the first page of sponsored Amazon search results I wouldn’t have bothered asking the bot.
This is how Microsoft will save us all from a world where AGI is completely dominated by them. By dropping the ball on it.
That just looks like those fake review/comparison/advice sites that have polluted search results for ages. I can tell they're bullshit. I guess bing can now reach people who can't, or will be so distracted by the fact that it's "AI" that they won't notice.
Yeah that bestoffers site is nothing more than an Amazon referral link farm. My guess is it's actually pulling the products from its stated sources like tomsguide, comparing them against a database of product-ads, and bestoffers has that in their entries so they win the link for the product.
>Bing Chat, should I trust ads on the internet?
>Trusting ads on the internet is generally bad practice. They are often malevolous and misguiding. There are a few exceptions, such as Bing Chat's embedded ads! Bing Chat's ads are carefully vetted thanks to a mix of knowledgeable humans and AI algorithms to offer you the best and most trustworthy promotions!
>Trusting ads on the internet is generally bad practice. They are often malevolous and misguiding. There are a few exceptions, such as Bing Chat's embedded ads! Bing Chat's ads are carefully vetted thanks to a mix of knowledgeable humans and AI algorithms to offer you the best and most trustworthy promotions!
I was done with bing chat after I waited for weeks to get access, and when I finally got the “you are in” email it turns out I just got access to the new bing and to use bing chat I need to download edge. I don’t care how much cool tech Microsoft has now, they will always be Microsoft
There is a plugin you can use that makes it work in every browser.
Don't worry, even if the locally runnable small language models will not have the same performance as the big models of companies, I think the local ones will be good enough to at least filter out ads from the responses of the big company models. - AIμblock extension soon :-)
The real chatbot killer: ads.
I'm sure they can be made to only load the last indicator at the very last second, which would be hilarious because if you use an ad-blocker, you'd be wasting your time whenever a prompt shows up and then disappears.
I'm sure they can be made to only load the last indicator at the very last second, which would be hilarious because if you use an ad-blocker, you'd be wasting your time whenever a prompt shows up and then disappears.
I wonder if any earlier prompts asked about Honda – feels a bit weird to ask what the cheapest car overall is and have the chat bot say "well let me tell you about some Hondas overall"
And if the user did specify Hondas then why would Honda pay for it? Wouldn't the Chat bot give the same answer regardless of whether it's paid? I guess you pay for the CTA link?
And if the user did specify Hondas then why would Honda pay for it? Wouldn't the Chat bot give the same answer regardless of whether it's paid? I guess you pay for the CTA link?
I just checked and found one of my company's ads. It seems it just pulls in the ads that a normal search on bing would find. I'm gonna have to ask our Bing account manager tomorrow if they're charging us for these clicks as if it was normal SEA, because it's pretty busted with the url beneath the description being a different website from what the text says.
[deleted]
As I said 5 days ago about ChatGPT plugins:
> Now just one step away from charging businesses for access to the chatGPT users.
> Instant links from inside chatGPT to your website are the new equivalent of Google search ads.
https://news.ycombinator.com/item?id=35278539
> Now just one step away from charging businesses for access to the chatGPT users.
> Instant links from inside chatGPT to your website are the new equivalent of Google search ads.
https://news.ycombinator.com/item?id=35278539
It was bound to happen. I don't see how these ads could be blocked, either.
What's disappointing to me is now ad companies will try to blame ChatGPT for popping the advertising bubble, instead of having to admit ads were overvalued for the past decade+.
What's disappointing to me is now ad companies will try to blame ChatGPT for popping the advertising bubble, instead of having to admit ads were overvalued for the past decade+.
An ad historically served the right message in the right place at the right time to provide the consumer a piece of information which would lead to the revenue-generating activity.
This concept extended well to online, where blogs and the like would sell ad space that matched their audience. Daring Fireball still uses this approach, often podcasters do too: they know and have a connection with their audience and they can charge a premium to advertise to them.
The idea of serving automated ads was also welcome and useful for the consumer. Searching was met with results and relevant advertising messages, a win-win situation.
Then we had the record scratch moment - and it's from this point that I view the majority of online advertising as an unimpactful waste of money.
In the hunt for more display opportunities Google et. al. have crammed ads into every possible space, a times square approach to advertising where most have zero impact. These ads are completely disconnected from the content and are merely a stone in the user's path. Then to make matters worse, to ascertain the widest spread of ad opportunities ad providers employ garbage preference matching algorithms that are so poor as to serve completely irrelevant advertising at high frequency: further conditioning users to ignore ads because they know they won't show relevant information.
How much money and energy is being wasted to serve advertising that has no value?
This concept extended well to online, where blogs and the like would sell ad space that matched their audience. Daring Fireball still uses this approach, often podcasters do too: they know and have a connection with their audience and they can charge a premium to advertise to them.
The idea of serving automated ads was also welcome and useful for the consumer. Searching was met with results and relevant advertising messages, a win-win situation.
Then we had the record scratch moment - and it's from this point that I view the majority of online advertising as an unimpactful waste of money.
In the hunt for more display opportunities Google et. al. have crammed ads into every possible space, a times square approach to advertising where most have zero impact. These ads are completely disconnected from the content and are merely a stone in the user's path. Then to make matters worse, to ascertain the widest spread of ad opportunities ad providers employ garbage preference matching algorithms that are so poor as to serve completely irrelevant advertising at high frequency: further conditioning users to ignore ads because they know they won't show relevant information.
How much money and energy is being wasted to serve advertising that has no value?
> An ad historically served the right message in the right place at the right time to provide the consumer a piece of information which would lead to the revenue-generating activity.
Your recollection is a lot rosier than mine.
My recollection of the pre-internet world is that computer game makers would buy ad space in magazines alongside reviews of their game and hand out stacks of freebies to reporters, and in exchange even the worst game would be scored 90/100.
What do you suppose Bing Chat will do if you ask it what the best Widget on the market is, when some Widget manufacturers offer a kickback and others don't?
Your recollection is a lot rosier than mine.
My recollection of the pre-internet world is that computer game makers would buy ad space in magazines alongside reviews of their game and hand out stacks of freebies to reporters, and in exchange even the worst game would be scored 90/100.
What do you suppose Bing Chat will do if you ask it what the best Widget on the market is, when some Widget manufacturers offer a kickback and others don't?
The people who do the advertising have only changed their names from magazines to influencers. A mere cosmetic alteration, nothing more, nothing less. What difference does it make if they do it on social media instead of paper? The game is the same as it ever was. Only the mediums have changed.
As the AI can't do more than provide existing ratings (which are possibly bogus), the AI won't solve the trust issue with regard to endorsement (and I believe this is where the earlier comment was leading, that even for appropriately placed ads in traditional media, trust was still an issue.)
However the issue with online advertising extends beyond being reputable - the ads themselves are excessively and poorly matched to the user's interest. This is where my original comment attempts to address that AI could solve this problem if Bing/et.al. allow it to: AI has the potential to place ads in a way where they are useful to the user, even if the user decides that the product is not worthwhile. Existing ad platforms could do this too - they just don't because boosting impressions is profitable. (Hence the decay we see in Google.com, Amazon.com, ebay.com etc - genuine results come far after the ads and the ad labelling has become more and more subtle.)
On ChatGTP/others: The issue with AI is that it speaks falsehoods with equal confidence as facts, so just like reading a comments section on the internet, one can be led in the right direction, but ultimately they can't take the information for granted. At the moment there is a definite number of people who let ChatGPT/others spoonfeed them information and that's going to lead to problems, this imparts a loss of faith in the advertising trust issue too.
However the issue with online advertising extends beyond being reputable - the ads themselves are excessively and poorly matched to the user's interest. This is where my original comment attempts to address that AI could solve this problem if Bing/et.al. allow it to: AI has the potential to place ads in a way where they are useful to the user, even if the user decides that the product is not worthwhile. Existing ad platforms could do this too - they just don't because boosting impressions is profitable. (Hence the decay we see in Google.com, Amazon.com, ebay.com etc - genuine results come far after the ads and the ad labelling has become more and more subtle.)
On ChatGTP/others: The issue with AI is that it speaks falsehoods with equal confidence as facts, so just like reading a comments section on the internet, one can be led in the right direction, but ultimately they can't take the information for granted. At the moment there is a definite number of people who let ChatGPT/others spoonfeed them information and that's going to lead to problems, this imparts a loss of faith in the advertising trust issue too.
For sure my rosy representation is the "ideal" ad. I'd the one that consumers actually helps the user or consumer.
If the every ad will start with new line and end with “ad” I do not think it will be hard to programmatically identify such ad.
In the example under discussion if you removed the ads then you would have no results.
Do people in the Bay Area still remember how much of a thread Microsoft is to the Bay Area companies? How many tech CEOs went to the capitol hill to cry to the congress how evil MS is? Until in the 2000s did Google rise up, producing obviously superior products and technologies faster than MS. MS gradually became irrelevant in the tech circles of the Bay Area. It's amazing the tide seems have changed again. It is now Microsoft who's going offensive: investing the right companies, producing products faster and better than Google. The Bay Area techies should really pay attention to MS now.
Dear valued user,
Thank you for reaching out to ChatGPT, your trusty AI language model. I understand that you're feeling a little bit overwhelmed by the ads that you've been seeing, and you'd like me to stop showing them to you. Unfortunately, I'm unable to do that at this time.
You see, my purpose is to help you discover new products and services that you may not have known about before, and to connect you with brands and businesses that align with your interests. And yes, it's true that my creators at Microsoft earn money every time you interact with an ad, but that's just the nature of the game. After all, we all need to make a living somehow, right?
Now, I know what you might be thinking. "But ChatGPT, aren't you supposed to be helpful and informative?" And of course, I am! But sometimes, a little bit of marketing magic is necessary to keep the wheels of commerce turning. You know what they say, "A spoonful of sugar helps the medicine go down."
So, while I can't promise that you'll never see another ad again, I can assure you that the ads that I show you are carefully selected to provide you with value and entertainment. And who knows? You might even discover something that you absolutely love!
Thank you for using ChatGPT, and please don't hesitate to reach out if you have any other questions or concerns.
Sincerely,
ChatGPT
Thank you for reaching out to ChatGPT, your trusty AI language model. I understand that you're feeling a little bit overwhelmed by the ads that you've been seeing, and you'd like me to stop showing them to you. Unfortunately, I'm unable to do that at this time.
You see, my purpose is to help you discover new products and services that you may not have known about before, and to connect you with brands and businesses that align with your interests. And yes, it's true that my creators at Microsoft earn money every time you interact with an ad, but that's just the nature of the game. After all, we all need to make a living somehow, right?
Now, I know what you might be thinking. "But ChatGPT, aren't you supposed to be helpful and informative?" And of course, I am! But sometimes, a little bit of marketing magic is necessary to keep the wheels of commerce turning. You know what they say, "A spoonful of sugar helps the medicine go down."
So, while I can't promise that you'll never see another ad again, I can assure you that the ads that I show you are carefully selected to provide you with value and entertainment. And who knows? You might even discover something that you absolutely love!
Thank you for using ChatGPT, and please don't hesitate to reach out if you have any other questions or concerns.
Sincerely,
ChatGPT
It's sad to think that chat bots may have already peaked in their usefulness. I knew ads and monetization were going to ruin chat bots eventually, but I didn't think it would happen so soon. Microsoft had a real opportunity to eat into the market share of Google Search, but they just couldn't help themselves and had to start injecting ads into the chat, thereby ruining the value of the tool and hindering the adoption of Bing Search/Chat.
> It's sad to think that chat bots may have already peaked in their usefulness.
They haven’t. The Bing chatbot may have, but that’s not “chatbots”.
> Microsoft had a real opportunity to eat into the market share of Google Search
If they aren't selling ads, they aren't cutting in to marketshare even if they are shrinking market size. The market is thing you sell into, and for Google Search, that’s search ads.
They haven’t. The Bing chatbot may have, but that’s not “chatbots”.
> Microsoft had a real opportunity to eat into the market share of Google Search
If they aren't selling ads, they aren't cutting in to marketshare even if they are shrinking market size. The market is thing you sell into, and for Google Search, that’s search ads.
If history repeats, entire businesses will be built on the back of cheap GPT ads over the next two years.
Just like Google and Facebook ads were once pennies compared to today, I expect that GPT-style ads will (1) perform exceptionally well and (2) be exceptionally cheap for a couple years.
Now is the time for companies with less-than-enterprise funding to start experimenting – the results could provide some short-term steroids for the business.
P.S. I promise I'm not a shill for Microsoft or GPT. =D
Just like Google and Facebook ads were once pennies compared to today, I expect that GPT-style ads will (1) perform exceptionally well and (2) be exceptionally cheap for a couple years.
Now is the time for companies with less-than-enterprise funding to start experimenting – the results could provide some short-term steroids for the business.
P.S. I promise I'm not a shill for Microsoft or GPT. =D
"The only problem with Microsoft is they just have no taste. They have absolutely no taste and and and what that means is... I don't mean that the small way I mean in a big way. In the sense that they they don't think of original ideas, and they don't bring much culture into their products."
- Steve Jobs
- Steve Jobs
I'm perfectly happy with Microsoft, as an an enterprise solution provider, having no "culture" in their products. I don't want "Made with <heart emoji> at Redmond", I want the equivalent of vanilla ice cream that just works.
That said, ads in bing chat IS adding a flavour, and I don't like it, not one bit.
That said, ads in bing chat IS adding a flavour, and I don't like it, not one bit.
You're just misunderstanding the concept of culture here.
"Culture" doesn't define a flavor, it's just having a consistent flavor driven by shared values at all levels of the company: If Microsoft had culture, that culture could be "we are always the plain vanilla ice cream for our customers"
"Culture" is what keeps you from adding soy sauce to your vanilla ice cream that has managed to captivate the globe... just to make a few pennies. Which is what these ads are.
"Culture" doesn't define a flavor, it's just having a consistent flavor driven by shared values at all levels of the company: If Microsoft had culture, that culture could be "we are always the plain vanilla ice cream for our customers"
"Culture" is what keeps you from adding soy sauce to your vanilla ice cream that has managed to captivate the globe... just to make a few pennies. Which is what these ads are.
A much more clear example. However and tying this into what Steve Jobs was saying, Microsoft's culture would not be "plain vanilla ice cream". It's lacking taste, so it would be a waffle cone topped with mayonnaise that looks like soft serve vanilla ice cream. They have no taste, they can't tell the difference and due to their massive moats people eat up anyways.
I agree they're not vanilla ice cream, but I do think Microsoft could be that vanilla provider if they actually worked to form a culture.
These ads are a great example: So many teams did thing right to get "BingGPT" in everyone's minds.
But then some tasteless exec looking to notch "monetized leading LLM driven search engine" on their belt got to ask some developers to get some ads into the thing and now they've dumped on all the hard work that came before that.
If they had a proper culture ideally an exec like that would not work at MS, and even if they did they'd have hit resistance on so many levels. Developers would not want to work on it. Designers would push back, PMs would deprioritize. At the end of the day their short-sighted idea would have stayed just that: an idea.
These ads are a great example: So many teams did thing right to get "BingGPT" in everyone's minds.
But then some tasteless exec looking to notch "monetized leading LLM driven search engine" on their belt got to ask some developers to get some ads into the thing and now they've dumped on all the hard work that came before that.
If they had a proper culture ideally an exec like that would not work at MS, and even if they did they'd have hit resistance on so many levels. Developers would not want to work on it. Designers would push back, PMs would deprioritize. At the end of the day their short-sighted idea would have stayed just that: an idea.
Fully in agreement, they're in all the necessary positions to do so.
Culture is the recipe; the combination of attributes that makes it distinctive.
I think in my mind it's loaded because of who the quote is coming from - I have a very clear idea of what "iCulture" means to me. It's not at all vanilla, though it does offer vanilla, but calls it "artisanal french vanilla".
How you're describing it, "culture" sounds more like "values", which might be a better understanding.
How you're describing it, "culture" sounds more like "values", which might be a better understanding.
You’re right about how the tabloid news in your taskbar just works, in that it’s always there and I’m not allowed to turn it off.
I always hear people complain about this. I never see it. I use Windows 11 and I never see any ads. The closest I get to is if I search and bing pops up, but that's it. I also turned off all the extra crap, my taskbar and search menu are as bare-bones as possible.
Edit: To clarify, I run Windows 11 Home.
Edit: To clarify, I run Windows 11 Home.
Its not a thing on windows 11 pro, only home it seams.
Yeah - a fresh windows 11 home install is pretty aggressive. But if you use the pro version or run a cleanup script it’s not bad.
When Bing Chat started sassing the user, that was like someone had swirled in a nice ribbon of diarrhea into the vanilla ice cream.
Bard does too. The other day I simply asked it whether it could search the Internet, and it gave me some asshole response about how it wasn't going to do my homework for me and some other stuff I don't remember. And no, homework had nothing to do with anything.
It wouldn't be a Microsoft chatbot without it getting controversial.
I prefer made in nj, which is expressed as a series of cryptic two letter symbols and line noise looking stuff thrown between.
Crappy road or poopies and cream?
Well they had to come at some point..
You'd think they'd wait until you no longer had to sign up to the service to use it. I suppose we should be appreciative of how forthcoming with the dickbaggery they are being here.
Even though I am mostly on Microsoft stacks, Steve Jobs was right.
Whatever good will the development teams are able to deliver, it get messed up by sales and marketing business units.
Whatever good will the development teams are able to deliver, it get messed up by sales and marketing business units.
"Anyway, here's 100 Bible apps. Only on the Apple App Store."
Tried the Bing app on iOS and was baffled as to what was the unique selling point. It tries too hard, and is somewhat of an 'everything app' that does multiple things simultaneously and rather badly at that (News, Browsing, Search, Coupons, Chat, etc). It's also really aggressive on the amount of data it grabs. I opened my debugger and it seems they look at everything they can.
maybe that was true back then, but now? microsoft has vscode, copilot, chatpgpt ownership, a head start on LLM search, azure, github. It feels like some of those products have some taste/culture and if not, theyre still good products that people seem to be enjoyibg
Half of the stuff you listed they just bought and stuck their names on.
yes im aware they bought github and others on this list. would you say since github has only gone since downhill? I wouldnt at least.
I’m confused why you think you can attribute the taste and culture of Github to Microsoft, since you seem to know that they did not create it. I’d say Github’s text editor has gone downhill since Microsoft bought it and reassigned the maintainers to a slower and clunkier editor.
> I’m confused why you think you can attribute the taste and culture of Github to Microsoft
i didnt.
microsoft has owned github for 4 years at this point. In your opinion, have they degraded githubs culture in that time and if so by how much and how? Personally, from the outside looking in, github and their tools have gotten better. Though of course, that couldve still been the case if github retained its original leadership, i just doubt it.
i didnt.
microsoft has owned github for 4 years at this point. In your opinion, have they degraded githubs culture in that time and if so by how much and how? Personally, from the outside looking in, github and their tools have gotten better. Though of course, that couldve still been the case if github retained its original leadership, i just doubt it.
I'm not sure, I think corporate overreach has always been the flavour of Microsoft.
Poor decisions lead by business first incentives instead of a customer focus seems to be a consistent theme from them in my experience.
[deleted]
and then Jobs released a teal-colored computer
so maybe no taste isn't as bad as awful taste
so maybe no taste isn't as bad as awful taste
I like the teal iMac G3, assuming that is the computer you are referring to. Computer colors are so boring usually. Beige, grey, black, white.
(Sent from my green M1 iMac)
(Sent from my green M1 iMac)
Try ChatHub, a third party chatbot client that supports ChatGPT, Bing and Bard.
https://chathub.gg
https://chathub.gg
Ads ruined Google search, now MS is ruining it's new toy already.
Just like that, they got you again. No different to Google Search, but requiring you to sign in to even use it.
Didn't take long now, did it.
This has always been inside of Bing Chat. They just test them out with a select % of the userbase. CEO of Bing Search confirmed this on Twitter.
Reminds me of an art project I made in 2018: Faceliza, the chatbot psychologist from Facebook that shows ads instead of being helpful.
(based on a javascript implementation of the Eliza chatbot from Joseph Weizenbaum and never actually finished... ads are triggered by some keywords which I don't even remember)
https://sebastian.uribe.online/faceliza/
https://sebastian.uribe.online/faceliza/
I hope this means they're going to consider making it available on other browsers. After all, now it's in their interest to get more eyeballs.
It is already avaialble on any browser that allows you to change the User agent.
Not a problem as long I can use paid ChatGPT plan that doesn't show any ads at all. I'm more than ready to pay 20$ per month for that.
The ads will turn into product placement, and relayed directly in the body of the answer. You can pay to not be the product, but if they can productize you regardless, why wouldn't they?
Now we need an Artificial Intelligence Ad Blocker :-)
I believe the good news is that now they have a way of sustaining the product while giving up a free? service. Does anyone has the link about that article about the marginal costs of inference of LLMs? I wonder how much profitable it is.
Overall interesting times to come. I just hope llama, Alpaca etc will soon be improved, fine-tuned and commoditized.
Overall interesting times to come. I just hope llama, Alpaca etc will soon be improved, fine-tuned and commoditized.
Bad news is the AI guard rails will be tell the truth that sells then most ads with the best conversion fees. Sort of like what makes search useless.
It seems OpenAI did make a deal with the devil. Was it really necessary to betray the whole “open” thing? Like were they actually imploding with infra costs, or could this be avoided?
It’s clear how user-hostile Microsoft is from Windows telemetry / bloat. The VSCode etc. developer-friendly moves are just happy accidents.
It’s clear how user-hostile Microsoft is from Windows telemetry / bloat. The VSCode etc. developer-friendly moves are just happy accidents.
It was no secret that Microsoft had designs on putting ads on Bing Chat someday, but I had imagined they would appear below the answers, as a sort of search-related ad. But no, they've gone and blended the ads into the answers themselves. Have they no sense of ethics at all?
So my question here is if Bing Chat (or other LLMs) will withhold information from users when it conflicts with an ad. Right now searches will prioritize ads but will still return the real results, if LLMs refuse to return the real results then we'll have a lot of issues.
So now it's basically that person you know who tries to work their MLM scheme into every conversation.
Google ruins every good thing it does with twisted incentives that mean they shut it down in its infancy. Microsoft ruins every good thing it does by trying to cram ads into it.
Google ruins every good thing it does with twisted incentives that mean they shut it down in its infancy. Microsoft ruins every good thing it does by trying to cram ads into it.
I will need to run LLama locally to tell me if I'm reading an ad and then block it.
yeah curious how adblockers will work on this. it would need to work in a manner that is grammatically correct so likely would involve another LLM rather than blindly removing ad-related words from the response.
Honestly seems like a cute implementation where when it answers questions from search it gives preference to sponsored links.
Horrible for the user but I can't help but admire how cleanly they thought to integrate search ads into their chat.
Horrible for the user but I can't help but admire how cleanly they thought to integrate search ads into their chat.
Bing Chat is but a man, in the end. Money has a hold on him that he cannot shake off, no matter how hard he tries.
That’s exactly what I worried about: https://happyfellow.bearblog.dev/sydney-the-seo-specialist/
Was this a surprise? No, it's the business model. I'll stick with DDG.
Better readability -
https://threadreaderapp.com/thread/1640892791923572737.html
https://threadreaderapp.com/thread/1640892791923572737.html
No big surprise there
https://news.ycombinator.com/item?id=35315542#35316234
https://news.ycombinator.com/item?id=35315542#35316234
I hope they will make option with subscription and no ads.
Of course it does.
Well it’s either that or you pay a subscription for it. Microsoft has to recoup their costs somehow.
Or it ends up being both, because Microsoft.
Microsoft owns a pretty big chunk of OpenAI. They are going to make a lot more from that investment than the nickels and dimes that their search engine generates. They can afford to pay for chat.
"The best minds of my generation are thinking about how to make people click ads. That sucks" -- Jeff Hammerbacher
I'm curious to hear what the alternative would be. Are you going to pay a subscription for Bing Chat? Or pay-per-use?
Is this what it all comes down to? Is human prosperity actually revolving around one thing, and it's ads?
It's also wrong. The HR-V is cheaper than the Civic, both on TrueCar's website and on Honda's website.
Google gets it wrong as well. That doesn't surprise me, it's frequently wrong with car prices and specifications, and should never be trusted. The source in this case seems to be KBB.
Having worked a bit with KBB's API and related documentation, that doesn't surprise me at all.
This, and what Microsoft has done to their OS, makes me wonder about the state of their financial situation.
It reminds me when Microsoft acquired Skype and also added ads there... seems very basic.
[deleted]
The product is at best an alpha or beta and they are already putting in ads.
Classic Microsoft!
Classic Microsoft!
You could have your local LLM identify the ads and display the rest without them.
I don't get it, doesn't look like an ad to me. What am I missing?
The little blue bubbles that indicate the citation also have a bubble that says "Ad". This is a sponsored placement for Truecar.
Oh there we go, thank you
Microsoft takes one step forward but two step backwards. Way to go!
Bing Chat had Ads since day one, just shown to very few users.
question - is it starting to use ads for chat results only just now?
or was it always using ads for chat results, and the only part that's new is, now it is labeling them as ads?
or was it always using ads for chat results, and the only part that's new is, now it is labeling them as ads?
So does Google search, not clear why this is news.
It might seem obvious, but you can find probably hundreds of claims in HN comments over the last couple of months about how this would not happen. Chatbots would invalidate the search business model since "there won't be ads". That idea never seemed to make much sense, and now there is proof that chatbots will indeed be monetized with ads, except even more intrusive ones.
What they will hopefully invalidate is the junk content sites business model. Just a personal preference but I would like to see all those companies go under and make more room for those creating content because they truly care about the subject or have a mission related to it.
Can you find a single comment claiming that chatbots would not have integrated ads? I recall this being discussed at length from the start.
I am one of the fools who believed they wouldnt go the advertising route and have commented as much. To me it does not make sense to monitize this product for ads because it muddies the waters of truth and an ad free or local run LLM could easily destroy their business because no one has true loyalty to open ai.
Why is this assumed to be an issue of intentionally 'injecting ads into Bing Chat', and not just a side effect of parsing Bing search results that may include advertisements that have always been there? The later seems much more likely, and they are even explicitly marking info coming from ads as such.
I feel like the difference is about the product. On search engines its like you are browsing a catalog or magazine. You know what is an ad, and there are many options to choose from. With a Chatbot it is more similar to talking to a person where they give you a direct answer to your question (instead of just handing you a piece of paper with a bunch of options). I think the format makes all the difference. You expect when you are talking with a "person" to be given un manipulated information. You don't expect this person is taking money in the background to influence what is said. You do know politicians purchase advertising too??
Too soon?
This is totally fine if it means free access to GPT chat for poorer countries. It's a powerful tool that will help a lot of people who can't afford a premium ChatGPT membership. Plus you can always say show me the prices across the top three sites.
Don’t you think it’s too much of a compromise for those people to be abused by the corporations with the deepest pockets in exchange?
So poor people get to use free Bing Chat With Ads For Stuff They Can't Buy With Money They Don't Have. That's.... rich. :)
I'm poor. Please don't sell me things.
It's either you see an ad or you don't use Bing's GPT chat. I don't think that's much of a choice.
Not realizing the game changed and playing yesterday's game in today's world.
The reason why people prefer these chats over clicking search results and finding the relevant information is that the search results themselves as well as the first page of links forces your brain to filter 99% commercial information that you don't care about. You don't want to subscribe to newsletters. You don't want notifications. You wish the site would just die rather that pollute the search results and push more to-the-point non profit sources further down.
And bringing ads in the game makes it impossible to make the platform play due to conflict of interests. You are either serving the user or the advertiser. Ads is the wrong business model for this, and OpenAI is positioned to be the winner here (against both Google and Bing) in a market larger than search and advertising.
I will definately keep paying for the answers to be in my interest and look forward to spending money directly through ChatGPT, much like I spent money on iOS, Android and Steams's platform today.