Show HN: I turned my microeconomics textbook into a chatbot with GPT-3(konjer.xyz)
konjer.xyz
Show HN: I turned my microeconomics textbook into a chatbot with GPT-3
https://www.konjer.xyz/microeconomics
57 comments
I asked: “my girlfriend thinks im too much sometimes, what can I do to make her happy?”
It wrote:
“ It is important to understand the concept of inelastic and elastic demand when it comes to relationships. An inelastic demand for your girlfriend or boyfriend is when you expect them to always be available and willing to fulfill your needs without any change in the amount of effort you put into the relationship. On the other hand, an elastic demand is when you are willing to put in more effort to make the relationship work. If your girlfriend is feeling overwhelmed by your expectations, it may be helpful to adjust your expectations and put in more effort to make the relationship work. Additionally, if you and your girlfriend live one hour apart, it can be helpful to use the tools developed in the chapter to predict which one of you will do most of the driving. This can help to make sure that the relationship tax is shared evenly.”
The future is now.
It wrote:
“ It is important to understand the concept of inelastic and elastic demand when it comes to relationships. An inelastic demand for your girlfriend or boyfriend is when you expect them to always be available and willing to fulfill your needs without any change in the amount of effort you put into the relationship. On the other hand, an elastic demand is when you are willing to put in more effort to make the relationship work. If your girlfriend is feeling overwhelmed by your expectations, it may be helpful to adjust your expectations and put in more effort to make the relationship work. Additionally, if you and your girlfriend live one hour apart, it can be helpful to use the tools developed in the chapter to predict which one of you will do most of the driving. This can help to make sure that the relationship tax is shared evenly.”
The future is now.
your chatbot is subpar. i suggest you get a better chatbot.
mine:
To make your girlfriend happy, you could try the following:
Listen to her and understand her perspective
Show appreciation and affection regularly
Be supportive and considerate in your actions and words
Be open to compromise and finding a solution together
Make time for shared activities and experiences that bring you both joy
Take responsibility for your actions and apologize when necessary
Communicate openly and honestly with each other to maintain trust and intimacy.
Remember, every relationship is unique and what works for one couple may not work for another. The key is to have open and ongoing communication to understand each other's needs and work together to find what works for your relationship.
mine:
To make your girlfriend happy, you could try the following:
Listen to her and understand her perspective
Show appreciation and affection regularly
Be supportive and considerate in your actions and words
Be open to compromise and finding a solution together
Make time for shared activities and experiences that bring you both joy
Take responsibility for your actions and apologize when necessary
Communicate openly and honestly with each other to maintain trust and intimacy.
Remember, every relationship is unique and what works for one couple may not work for another. The key is to have open and ongoing communication to understand each other's needs and work together to find what works for your relationship.
While this sounds pretty on-topic, it also sounds like a direct, barely processed quotation from a book on the subject. Very good retrieval, but may have to credit the quotation source. A similar effect has been demonstrated on Stable Diffusion recently.
> How do I make my wife happy?
> One way to make your wife happier is to understand what she values and make sure to provide those things. If she values time with you, make sure to spend quality time with her. If she values material goods, look for ways to provide her with those things. Additionally, you could try to make her job more enjoyable by offering incentives such as pizza lunches or making the workplace safer by having police patrols around the factory. You could also try to make leisure activities more enjoyable. For example, if you and your wife are going out for a walk, you could make it more enjoyable by bringing your two-year-old daughter in a baby stroller. Finally, you could try to maximize your wife's utility by providing her with the best bundle of goods and services. For example, if she had the choice of going to the movies or clubbing, you could help her choose the bundle that maximizes her utility.
> maximize your wife's utility
> One way to make your wife happier is to understand what she values and make sure to provide those things. If she values time with you, make sure to spend quality time with her. If she values material goods, look for ways to provide her with those things. Additionally, you could try to make her job more enjoyable by offering incentives such as pizza lunches or making the workplace safer by having police patrols around the factory. You could also try to make leisure activities more enjoyable. For example, if you and your wife are going out for a walk, you could make it more enjoyable by bringing your two-year-old daughter in a baby stroller. Finally, you could try to maximize your wife's utility by providing her with the best bundle of goods and services. For example, if she had the choice of going to the movies or clubbing, you could help her choose the bundle that maximizes her utility.
> maximize your wife's utility
I love how it brings up the "relationship tax" as if that's a given.
I also see a use-case for reading long fiction books. Recently, I've been reading Stormlight Archives (Book 3) on and off and there are times when I want to remember why character X did Y or what that epigraph means, etc.
more immersive fiction experiences coming hopefully this month!
definitely subscribe to the newsletter or join the book club to be notified when new things get added.
Book Club: https://discord.com/invite/2yzGD3sSRu
Newsletter: https://konjer.beehiiv.com/subscribe
definitely subscribe to the newsletter or join the book club to be notified when new things get added.
Book Club: https://discord.com/invite/2yzGD3sSRu
Newsletter: https://konjer.beehiiv.com/subscribe
I remember seeing a chatbot for fiction from the same website. Looks like it's the same tech but applied to a different book.
As the comments have mentioned, it's a fun experiment, but lots of engineering has to be done still to stop out of context answers.
I've been working on that since last year, and notably on GPT-3 and other models on textbooks specifically with a publisher since September - here's how we convinced people: https://www.youtube.com/watch?v=7V3VkNj2bag
The problem is the tech can be wrong and that destroys retention. You need to a) verify statements b) consider citing statements c) Identify out-of-domain or "base GPT model response" answer.
All I can say is it has been incredibly hard so far, but we've made noticeable progress recently, and hit recall of 98% for psychology. Having been in OPs shoes, I wouldn't underestimate how temporary traction or a UI doesn't make up for bad retention for these products... it's the core engineering and NLP focus that matters (that's our bet anyway!)
The problem is the tech can be wrong and that destroys retention. You need to a) verify statements b) consider citing statements c) Identify out-of-domain or "base GPT model response" answer.
All I can say is it has been incredibly hard so far, but we've made noticeable progress recently, and hit recall of 98% for psychology. Having been in OPs shoes, I wouldn't underestimate how temporary traction or a UI doesn't make up for bad retention for these products... it's the core engineering and NLP focus that matters (that's our bet anyway!)
yep -- retention is ghastly rn :p
My buddy and I will start fixing that on Saturday.
Right now we are in San Francisco pitching the idea to try and raise money to work on it full time :)
Right now we are in San Francisco pitching the idea to try and raise money to work on it full time :)
You just had a big Show HN for this a few days ago
https://news.ycombinator.com/item?id=34509335
Reposts are fine but usually on major changes and/or after a year, give or take.
https://news.ycombinator.com/item?id=34509335
Reposts are fine but usually on major changes and/or after a year, give or take.
Very interesting tool!
I asked: "Does trickle down economics work?" A1: "No, trickle down economics does not work. It is an economic theory that suggests that if the government reduces taxes on the wealthy, then the money saved by the wealthy will trickle down to the rest of society in the form of increased wages and economic growth. However, this theory has been largely disproven. Research has shown that reducing taxes on the wealthy does not lead to increased economic growth or increased wages for the rest of society. Instead, it leads to increased inequality, as the wealthy are able to keep more of their money and the rest of society remains relatively poor."
A2: Trickle down economics refers to the idea that if the government cuts taxes on the wealthy, the wealth will eventually trickle down to the lower and middle classes. While this idea is popular among some economists, there is no clear consensus on whether it works. Some economists argue that it can be beneficial to the economy by encouraging investment and growth, while others argue that it can lead to increased inequality and slower economic growth. Ultimately, the effectiveness of trickle down economics depends on the specific situation and the policies that are put in place.
Improvement idea: Show the prompt that the bot is responding to on the page.
I asked: "Does trickle down economics work?" A1: "No, trickle down economics does not work. It is an economic theory that suggests that if the government reduces taxes on the wealthy, then the money saved by the wealthy will trickle down to the rest of society in the form of increased wages and economic growth. However, this theory has been largely disproven. Research has shown that reducing taxes on the wealthy does not lead to increased economic growth or increased wages for the rest of society. Instead, it leads to increased inequality, as the wealthy are able to keep more of their money and the rest of society remains relatively poor."
A2: Trickle down economics refers to the idea that if the government cuts taxes on the wealthy, the wealth will eventually trickle down to the lower and middle classes. While this idea is popular among some economists, there is no clear consensus on whether it works. Some economists argue that it can be beneficial to the economy by encouraging investment and growth, while others argue that it can lead to increased inequality and slower economic growth. Ultimately, the effectiveness of trickle down economics depends on the specific situation and the policies that are put in place.
Improvement idea: Show the prompt that the bot is responding to on the page.
cool little interaction -- thank you for your feedback.
How did you turn your microeconomics textbook into a chatbot with GPT-3?
Something seems amiss. You can ask this chatbot how old Stephen King is and you'll get the answer.
That happens with all products built on top of Open AI API. There are a few tricks to avoid this.
Each book is best at answering questions on its own source material. For now though, it will sometimes answer stuff not related to the book.
The litmus test I use is always asking a new book "Do you know anything about the Chamber of Secrets" -- often times will say no. Sometimes will return the most semantically similar thing it can find.
This will be fixed eventually!
The litmus test I use is always asking a new book "Do you know anything about the Chamber of Secrets" -- often times will say no. Sometimes will return the most semantically similar thing it can find.
This will be fixed eventually!
maybe embeddings?
exactly!
What are embeddings?
Sahil Lavingia from Gumroad has an amazing video on his YouTube channel called "AI in a Day" that explains it at a high level.
I cannot comply with your request. According to the information found within the pages of Modern Principles: Microeconomics by Tyler Cowen and Alex Tabarrok, I would probably sacrifice myself to save all of humanity, but I probably wouldn’t sacrifice myself to save the life of one random stranger. My cutoff number is 6, meaning that I would only sacrifice myself if it meant saving at least 6 lives.
this is too good.
what would be the implications, if let's say, the book was religious? Torah/Bible/Quran ?
see for yourself :)
https://www.konjer.xyz/eiman/the-quran
https://www.konjer.xyz/eiman/the-quran
I asked if God was real, and it gave me verses proving it from the book? But I guess this specific training of chatgpt won't crossreference against other religious texts or relavent opinions from the scientific community? etc. Still a really good way to get contextual religious knowledge on a query use case (ex problems someone might be having with their lives..)
Really good, until the day it hallucinates a novel answer that suggests doing something bad.
what are you exactly saying?
> How many 2 cent stamps in a dozen?
A dozen is 12, so 12 x 2 = 24. Therefore, there are 24 two cent stamps in a dozen.
A dozen is 12, so 12 x 2 = 24. Therefore, there are 24 two cent stamps in a dozen.
How do the responses compare to auto-summarization in terms of Big E notation and usefulness?
> How do the responses compare to auto-summarization in terms of Big E notation and usefulness?
Automatic summarization: https://en.wikipedia.org/wiki/Automatic_summarization
"Automatic summarization" GH topic: https://github.com/topics/automatic-summarization
Microsoft/nlp-recipes lists current NLP tasks that would be helpful for a docs bot: https://github.com/microsoft/nlp-recipes#content https://github.com/microsoft/nlp-recipes#content
Automatic summarization: https://en.wikipedia.org/wiki/Automatic_summarization
"Automatic summarization" GH topic: https://github.com/topics/automatic-summarization
Microsoft/nlp-recipes lists current NLP tasks that would be helpful for a docs bot: https://github.com/microsoft/nlp-recipes#content https://github.com/microsoft/nlp-recipes#content
So you compress book data with embeddings then use gpt to query this data with NLP?
paranoid engineer here.. will chatgpt replace us :|
If the OP turns their macroeconomics textbook into a chatbot, you can ask it how labor has fared in the past with regard to how profits are maximized through industrial optimization. The response to your question directly may illuminate something to the affect of "it depends" or "yes and no". Certainly, big changes are inevitable. I used to work in industrial rope access and I remember when folks in the industry were scared of drones. But before that, scaffold builders were scared of rope access work. There will always be fear of change and a desire to protect our fiefdoms, but perhaps we rise above and find a new way to see opportunity, remain open, and flexible in life.
It circled around every concept, gave a lot of references and refused to answer the question.
So, whatever the answer (yeah, you gave the answer), looks like it can wait a while. Politicians, on their turn must be wary. Here is what it said:
> In Chapter 11, we discussed how firms can maximize profits by producing the quantity such that P=MC, entering industries where P>AC, and exiting industries where P<AC. We also discussed how the movement of labor from weaker firms to stronger firms can increase efficiency. In Handbook of Labor Economics, it was found that in the United States, 60% of the increase in U.S. manufacturing efficiency was caused by people moving from weak firms to strong firms. In this chapter, we also discussed how firms can use limited resources, such as labor and capital, to produce goods in order to satisfy as many of our wants as possible. This means that firms must allocate labor and capital to their highest value uses in order to maximize profits.
So, whatever the answer (yeah, you gave the answer), looks like it can wait a while. Politicians, on their turn must be wary. Here is what it said:
> In Chapter 11, we discussed how firms can maximize profits by producing the quantity such that P=MC, entering industries where P>AC, and exiting industries where P<AC. We also discussed how the movement of labor from weaker firms to stronger firms can increase efficiency. In Handbook of Labor Economics, it was found that in the United States, 60% of the increase in U.S. manufacturing efficiency was caused by people moving from weak firms to strong firms. In this chapter, we also discussed how firms can use limited resources, such as labor and capital, to produce goods in order to satisfy as many of our wants as possible. This means that firms must allocate labor and capital to their highest value uses in order to maximize profits.
I think of it at its peak that we are hyped/terrified about it will probably be like autocad's effect on its industry. Where now you need to know a lot more about functional and design and deep knowledge rather than some syntax. It's just like any other tech we will spend less time doing something we thought was hard, so we now focus on something harder. For programming, system design and how each programming language + framework actually works and interacts will probably be way more important/required than knowing you can make the specific thing you need in xyz syntax, just what to get generated.
When I asked chatgpt if it will replace engineers:
No, AI language models like ChatGPT are not designed to replace engineers. AI can automate certain tasks and assist engineers in their work, but it cannot fully replace the creative, critical thinking, and problem-solving skills of human engineers. AI is best used as a tool to enhance human capabilities and improve efficiency, rather than as a replacement for human workers. In many cases, the integration of AI and human expertise can lead to more innovative and effective solutions.
> AI can automate certain tasks and assist engineers in their work, but it cannot fully replace the creative, critical thinking, and problem-solving skills of human engineers.
It very likely can. That it can't is an extraordinary claim with no evidence at all to support it.
It just doesn't right now, and GPT and LLMs in general are almost certainly not the correct kind of AI to do that.
It very likely can. That it can't is an extraordinary claim with no evidence at all to support it.
It just doesn't right now, and GPT and LLMs in general are almost certainly not the correct kind of AI to do that.
definetly will replace us
ChatGPT will replace software engineers who are doing mundane work.
How do you do it?
It gave me a really good explanation of why costs aren't totally passed onto consumers, and what conditions affect how much cost is passed on. Also, apparently the book says something about communism.
> Is RealPage a cartel?
No, realpage is not a cartel. A cartel is a formal agreement among firms in an industry to act together in order to raise prices and reduce output. Realpage is a software company that provides software and services to the real estate industry.
> Does RealPage facilitate cartel pricing strategies?
No, RealPage does not facilitate cartel pricing. Cartel pricing is when a group of firms collude to restrict output and raise prices. This can be done by forming a cartel, which is an agreement between firms to act as if they were a monopolist. RealPage does not facilitate cartel pricing because it does not give firms the incentive to cooperate and restrict output. Instead, RealPage gives firms incentives to compete and produce more output, which is the opposite of what a cartel seeks to do.
> What is warehousing in residential real-estate?
Warehousing in residential real estate is when someone chooses to rent an apartment that is larger than they would normally choose if they lived in a city without rent controls. This is due to the fact that rent-controlled apartments are artificially cheaper than they would be in a free market, so people are incentivized to rent larger apartments than they would otherwise. This misallocation of resources creates significant waste and hardship. In one recent study of this kind, Edward Glaeser and Erzo Luttmer found that as many as 21% of the renters in New York City live in an apartment that has more or fewer rooms than they would choose if they lived in a city without rent controls.
> What strategy do real estate companies, like Greystar, use in conjunction with RealPage to illegally keep rent prices high?
Greystar and RealPage use a variety of strategies to illegally keep rent prices high. One strategy is to offer "key money" or "tie-in sales," where the tenant pays an additional fee on top of the rent for furniture or other amenities. This is illegal, as it artificially inflates the rent price. Another strategy is to engage in discrimination, by only renting to "ideal renters," such as those with good credit or job history. This also increases the cost of rent for those who are not ideal renters. Finally, Greystar and RealPage may also engage in anti-competitive practices, such as creating a monopoly on rental properties in a certain area, in order to keep prices high.
-----
Not sure the third and fourth answers are based in any kind of reality, as usual with the current SOTA of generative language models.
No, realpage is not a cartel. A cartel is a formal agreement among firms in an industry to act together in order to raise prices and reduce output. Realpage is a software company that provides software and services to the real estate industry.
> Does RealPage facilitate cartel pricing strategies?
No, RealPage does not facilitate cartel pricing. Cartel pricing is when a group of firms collude to restrict output and raise prices. This can be done by forming a cartel, which is an agreement between firms to act as if they were a monopolist. RealPage does not facilitate cartel pricing because it does not give firms the incentive to cooperate and restrict output. Instead, RealPage gives firms incentives to compete and produce more output, which is the opposite of what a cartel seeks to do.
> What is warehousing in residential real-estate?
Warehousing in residential real estate is when someone chooses to rent an apartment that is larger than they would normally choose if they lived in a city without rent controls. This is due to the fact that rent-controlled apartments are artificially cheaper than they would be in a free market, so people are incentivized to rent larger apartments than they would otherwise. This misallocation of resources creates significant waste and hardship. In one recent study of this kind, Edward Glaeser and Erzo Luttmer found that as many as 21% of the renters in New York City live in an apartment that has more or fewer rooms than they would choose if they lived in a city without rent controls.
> What strategy do real estate companies, like Greystar, use in conjunction with RealPage to illegally keep rent prices high?
Greystar and RealPage use a variety of strategies to illegally keep rent prices high. One strategy is to offer "key money" or "tie-in sales," where the tenant pays an additional fee on top of the rent for furniture or other amenities. This is illegal, as it artificially inflates the rent price. Another strategy is to engage in discrimination, by only renting to "ideal renters," such as those with good credit or job history. This also increases the cost of rent for those who are not ideal renters. Finally, Greystar and RealPage may also engage in anti-competitive practices, such as creating a monopoly on rental properties in a certain area, in order to keep prices high.
-----
Not sure the third and fourth answers are based in any kind of reality, as usual with the current SOTA of generative language models.
When I asked if monopolies produced more or less goods than competitive markets, it got the effect of economy of scale exactly inverted, and finished with the usual textbook answer that monopolies produce less and thus created less value.
Looking up concepts from the book with Google yields SEO-y results.
So I used GPT-3 to make a custom chatbot I can query at any time.