Ask HN: Are you using Google less lately because of ChatGPT?
10 comments
My search behaviour hasn't changed. I find chatGTP's answers to be too verbose, imprecise and the lack of visuals and links in its answers reduces utility for the domains I'm interested in (I think SEO works well in those domains). I often depend on substantiating my work with sources, so I need links to blogs, articles and research papers.
same here: I only use Google for links; I don't want general information from them (or anyone else).
Slightly. They're different kinds of searches.
ChatGPT is better at telling me what a block of code does or what my medical report says. Google for answering quick questions that don't need a full sentence.
For all the criticism ChatGPT gets, Google is also highly inaccurate.
I've wired GPT-3 to my command prompt, and I tend to use that for quick queries, like keyboard shortcuts and the command prompt to delete a file. It shouldn't replace document lookup tools like Dash though.
ChatGPT is better at telling me what a block of code does or what my medical report says. Google for answering quick questions that don't need a full sentence.
For all the criticism ChatGPT gets, Google is also highly inaccurate.
I've wired GPT-3 to my command prompt, and I tend to use that for quick queries, like keyboard shortcuts and the command prompt to delete a file. It shouldn't replace document lookup tools like Dash though.
It's a different sort of thing I'm after when I'm querying ChatGPT. I use ChatGPT if I have writers block and need it to draft a certain kind of email for me, but I can't believe a single thing it says so its usefulness is quite limited. Beyond that, it's a very good toy for me right now, but I'm sure it'll only improve from here.
Absolutely not. I haven't even used ChatGPT. I'm not against it, I just like to run things locally. I have gpt-2 and gpt-j and tons of different stable diffusion models local and I get way more enjoyment from diffusion rather than some large text prediction model.
All these models know are what they are trained on. That's it. What they are trained on along with whatever limited window of memory that they have for an open-ended chat type query.
I don't consider any large language model a replacement for research and searching. Although I do get a kick out of it for generating ridiculous nonsense and that brings me great joy because I get to laugh at it. Making nonsense top 10 lists or nonsense drink recipes. Those are hilarious times. Deliberately tuning the model so it's running a fever is just wonderful.
All these models know are what they are trained on. That's it. What they are trained on along with whatever limited window of memory that they have for an open-ended chat type query.
I don't consider any large language model a replacement for research and searching. Although I do get a kick out of it for generating ridiculous nonsense and that brings me great joy because I get to laugh at it. Making nonsense top 10 lists or nonsense drink recipes. Those are hilarious times. Deliberately tuning the model so it's running a fever is just wonderful.
Why bother responding to this question if you haven't actually used it? that's like not reading the article before commenting and surely you're better than that, right?
Because everyone else is using it and I've seen what it could do. I've used other large language models. I don't need to actually use it myself firsthand because I've observed millions of people using it secondhand. It's amazing for them but I just seen novelty in it and I don't see utility in it. I see utility in GitHub co-pilot. I see utility and being able to run gptj locally and then feeding that into visual studio code so I could run my own local GitHub copilot. I don't see any utility in a gated, run on someone else's computers, large language model.
People talk about monetizing it and somehow running apps off of it and all of this crap and all it does is benefit the creators of it. You will own nothing of it. You aren't running it on a computer that you own. You're going to be renting compute from them to run the thing that they made in hopes that you can make money off of it. It's silly! You're never going to come out on top in that proposition unless you can run it on hardware that you own and not just in the cloud because you're renting that hardware.
People talk about monetizing it and somehow running apps off of it and all of this crap and all it does is benefit the creators of it. You will own nothing of it. You aren't running it on a computer that you own. You're going to be renting compute from them to run the thing that they made in hopes that you can make money off of it. It's silly! You're never going to come out on top in that proposition unless you can run it on hardware that you own and not just in the cloud because you're renting that hardware.
You can read all you want about riding a bike, but there's no substitute for actually riding one. We're not even talking about something as dangerous as riding a bike!
The birth of Google in 1998 also brought about the birth of the SEO industry. 25 years later, there's still an SEO industry despite being built on top of something they don't have any control over. The beauty of a market economy is that it allows for people monetizing something they own nothing of, and that's totally fine!
The birth of Google in 1998 also brought about the birth of the SEO industry. 25 years later, there's still an SEO industry despite being built on top of something they don't have any control over. The beauty of a market economy is that it allows for people monetizing something they own nothing of, and that's totally fine!
I'm using Google less lately because of DuckDuckGo.
i certainly am