9 ms·

Using GPT-3 to pathfind in random graphs

165 points by tibbar 4y ago

speedgoose 4y ago
Have you tried to create a fine tuned model to see if it helps?
- ALittleLight 4y ago
  I don't believe the weights for GPT-3 are public and it would be expensive to fine tune such a large model.
  - dwohnitmok 4y ago
    OpenAI provides an API for fine-tuning. https://beta.openai.com/docs/guides/fine-tuning https://beta.openai.com/docs/guides/fine-tuning
  - speedgoose 4y ago
    I paid a bit more than $1 to fine tune the biggest GPT3 model for my use case. I didn’t push a lot of data and the fine tuning took 35 minutes (without counting the time waiting in the queue).
  - Der_Einzige 4y ago
    It's not that expensive to find tune huge models. You can find tune gpt-J or gpt-neo with a 3090
ALittleLight 4y ago
I find it a little unsettling that GPT-3 is able to do this well. This kind of problem is abstract and I assume its training data doesn't extensively cover this kind of thing. I wonder if it would work meaningfully worse if you used different words in the prompt other "edge" and "node" which are probably used in examples somewhat like this. Example prompt - Every turn each gloober flomps the same gloobers. If a yellow gloober flomps a blue gloober then the blue gloober turns yellow, otherwise, gloobers stay the same color. 1. gloober Roy flomps gloober Zoo and gloober Bat 2. gloober Zoo flomps gloober Bat and gloober Crystal 3. gloober Crystal flomps gloober X-Ray 4. gloober Bat flomps gloober Blue If gloober Zoo is the only gloober that is yellow, and all other gloobers are blue, will gloober Blue turn yellow, and, if so, at what turn? Answer: gloober Blue
- dwohnitmok 4y ago
  GPT-3 didn't even need the `gloober Blue` prompt as part of `Answer:`. It 0-shotted (no other training examples required) it straight out of the box. Prompt: Every turn each gloober flomps the same gloobers. If a yellow gloober flomps a blue gloober then the blue gloober turns yellow, otherwise, gloobers stay the same color. 1. gloober Roy flomps gloober Zoo and gloober Bat 2. gloober Zoo flomps gloober Bat and gloober Crystal 3. gloober Crystal flomps gloober X-Ray 4. gloober Bat flomps gloober Blue If gloober Zoo is the only gloober that is yellow, and all other gloobers are blue, will gloober Blue turn yellow, and, if so, at what turn? Answer: Continuation: gloober Blue will turn yellow on turn 4. > I find it a little unsettling that GPT-3 is able to do this well. I agree and think that this is definitely the right reaction.
  - mbil 4y ago
    Reminds me of a paper that I saw the other day: "On the Paradox of Learning to Reason from Data"[0] [0]: http://starai.cs.ucla.edu/papers/ZhangArxiv22.pdf http://starai.cs.ucla.edu/papers/ZhangArxiv22.pdf
    - foota 4y ago
      This is fantastic, thanks for sharing!
  - nikkwong 4y ago
    I know that GPT-3 is not AGI nor modeled properly to ever represent what we would consider to be AGI; but this feels freakishly AGI-ish. This feels like the model is 'reasoning' in a way that I would not expect based on my intuition of how GPT-3 works. As it's been explained to me in lay terms; GPT is great at predicting the next word based on a chunk of previous context. This feels like it's doing much more than that as it's understanding the context and parameters that relate to a question embedded in the text, which seems unrelated to the original instruction of GPT-3 as I understand. Can anyone explain?
    - 4y ago
mjburgess 4y ago
The problem with research conducted in this way is that there's no theory of operation of the model, except the implicit one "it works like how I expect". And then there's no hypothesis test against predictions made based on this theory, ie., no one tries to disconfirm the model works that way. It is usually trivial to disconfirm. ML models almost always work some other way, usually exploiting coincidental statistical information in the training set. This turns the whole show into a superstitious activity, in which we think google answers our "local restaurants" query because it knows where I live (rather than say, using an IP location or one of 100 other methods). If you present a ML model as "capable of x", you're required to disprove that claim as best you can, and present the occasions where it failed.
- xyzzy123 4y ago
  It might not be rigorous science but it's a cracker of a blog post. It shows a surprising ability of language models, code is provided and plenty of mystery is left for the reader... which can be investigated by almost anyone for tens of dollars.
  - mjburgess 4y ago
    But is it an ability? I can't tell. What would a useless "random" algorithm find operating on these graphs as inputs? What would a useful one find? What would that be "useful" for? Is this system much closer to the useless or the useful? We cant tell. Without any explanation, we're left reading this blog post as a magic trick in the literal sense. Consider the rate at which it "uses edges that don't exist" -- this makes it not a graph traversal algorithm. So it is just exploting syntax in the input whose representation has some "graphlike" qualities which can be statistically exploited. Why is that interesting? I dont know. Is this a good system?
    - xyzzy123 4y ago
      I don't know! That's why it's fun. (EDIT: I can safely say this isn't practically useful... but it's interesting in the same way as watching a cat try to fill out a tax return. I'm not surprised the cat does it badly, I'm surprised that it seems to be doing it at all, even if it's really a trick).
taberiand 4y ago
The performance of this model using English descriptions suggests to me the possibility of a model that is built to work using precise mathematical notation performing even better. Are there models that are trained to receive and respond with a purely mathematical description of the problem?
- phi0 4y ago
  OpenAI has trained a model to solve Math Olympiad-type problems[1], with some initial success, but currently the model isn't good at coming up with proofs involving more than a couple arguments chained together. Still some very impressive work. [1] https://analyticsindiamag.com/openais-neural-theorem-prover-can-solve-math-olympiad-problems/ https://analyticsindiamag.com/openais-neural-theorem-prover-...
- lpasselin 4y ago
  Check out Google Minerva. https://ai.googleblog.com/2022/06/minerva-solving-quantitative-reasoning.html https://ai.googleblog.com/2022/06/minerva-solving-quantitati...
dr_dshiv 4y ago
Propositions: 1. GPT3 is AGI 2. In combination with people, GPT3 is superintelligence Strong claims. Anyone care to argue for or against?
- jmfldn 4y ago
  There is zero evidence for the AGI claim or even a weaker intelligence claim. GPT 3 is just a search engine on steroids. I'm regularly suprised by the results but it's honestly just an illusion as far as 'intelligence' goes. This has nothing to do with intelligence, and anyone who thinks otherwise is succumbing to the Eliza effect.
  - hackerlight 4y ago
    Why can't the same thing be said of human intelligence? If we knew more about neuroscience it'd probably seem just as trivial. Wetware trained on examples we've seen. Experience encoding neurons. Neurons running on the laws of physics. Data goes in, endogenous processing happens, then outputs. I don't see why it's categorically different. ANNs are just demystified because we made them and know how they work.
    - jmfldn 4y ago
      You've answered your own question. "why can't the same thing be said of human intelligence" "if we knew more about neuroscience..." So that's why we can't say the same thing. Fwiw, my take is that whatever the brain does, its a machine of some sort but possibly not even a turing machine. Many disagree but noone can prove me wrong. Yet at least.
      - pennaMan 4y ago
        So are we intelligence or just a pattern recognition system on steroids?
        jmfldn 4y ago
bilsbie 4y ago
Where’s the best place to try out one of these language models right now? Doesn’t have to be gpt3.
- theblazehen 4y ago
  Try signing up for https://beta.openai.com/playground https://beta.openai.com/playground, I got access to GPT-3 in a couple days. For GPT-2 you can try out https://transformer.huggingface.co/ https://transformer.huggingface.co/
jcims 4y ago
Would be interesting to see if it does better if the process was broken down into ’moves’ and the model asked/reminded what its next possible steps are along the way.
jokethrowaway 4y ago
I can't wait for the next version which will surely know how to pass a leetcode interview.
rytill 4y ago
Do other large language models perform well on this task, e.g. EleutherAI models?
ibestvina 4y ago
Great writeup! I made a couple of changes in your code and it seems it gets ~90% of examples correct (inside the first three categories you have) with aroung 65% having either optimal path or correctly saying that there is no path. What I changed was: 1. Used code-davinci-002 (codex) 2. Instead of using an explanation of how these tasks work, I changed your code prompts slightly and instead gave it 3 examples (two with paths, and one to show it what to output when there is no path present). 3. Changed the output so that it has to tell me which edges, instead of which nodes, it is traversing (this helps GPT to avoid using nonexistent edges it seems). Here is an example: https://pastebin.com/D7Hn95VC https://pastebin.com/D7Hn95VC Last "Problem" is the real problem we want to solve of course, everything else is static. I'll post my code when I finish testing :) EDIT: I used Codex not because I think it's better suited for this, but because it's currently free. I don't have enough credit ATM to run 1000 iterations with text-davinci, so I have no idea what difference using Codex made.
- tibbar 4y ago
  This looks amazing, I’m looking forward to seeing your version!
  - ibestvina 4y ago
    Here are the stats :) - Found the optimal path 504 - Found a (non-optimal) path 293 - Correctly reported no path exists 131 - Found a solution when none existed 19 - Ended on wrong node 0 - Used edges that don't exist 53 - Started with the wrong node 0 - Incorrectly reported no path exists 0 - Total: 1000 Here is the code: https://github.com/ibestvina/gpt3-graph-search https://github.com/ibestvina/gpt3-graph-search

165 points by tibbar 4y ago

speedgoose 4y ago

Have you tried to create a fine tuned model to see if it helps?

ALittleLight 4y ago
I don't believe the weights for GPT-3 are public and it would be expensive to fine tune such a large model.
- dwohnitmok 4y ago
  OpenAI provides an API for fine-tuning. https://beta.openai.com/docs/guides/fine-tuning https://beta.openai.com/docs/guides/fine-tuning
- speedgoose 4y ago
  I paid a bit more than $1 to fine tune the biggest GPT3 model for my use case. I didn’t push a lot of data and the fine tuning took 35 minutes (without counting the time waiting in the queue).
- Der_Einzige 4y ago
  It's not that expensive to find tune huge models. You can find tune gpt-J or gpt-neo with a 3090

ALittleLight 4y ago

I find it a little unsettling that GPT-3 is able to do this well. This kind of problem is abstract and I assume its training data doesn't extensively cover this kind of thing. I wonder if it would work meaningfully worse if you used different words in the prompt other "edge" and "node" which are probably used in examples somewhat like this. Example prompt - Every turn each gloober flomps the same gloobers. If a yellow gloober flomps a blue gloober then the blue gloober turns yellow, otherwise, gloobers stay the same color. 1. gloober Roy flomps gloober Zoo and gloober Bat 2. gloober Zoo flomps gloober Bat and gloober Crystal 3. gloober Crystal flomps gloober X-Ray 4. gloober Bat flomps gloober Blue If gloober Zoo is the only gloober that is yellow, and all other gloobers are blue, will gloober Blue turn yellow, and, if so, at what turn? Answer: gloober Blue

dwohnitmok 4y ago
GPT-3 didn't even need the `gloober Blue` prompt as part of `Answer:`. It 0-shotted (no other training examples required) it straight out of the box. Prompt: Every turn each gloober flomps the same gloobers. If a yellow gloober flomps a blue gloober then the blue gloober turns yellow, otherwise, gloobers stay the same color. 1. gloober Roy flomps gloober Zoo and gloober Bat 2. gloober Zoo flomps gloober Bat and gloober Crystal 3. gloober Crystal flomps gloober X-Ray 4. gloober Bat flomps gloober Blue If gloober Zoo is the only gloober that is yellow, and all other gloobers are blue, will gloober Blue turn yellow, and, if so, at what turn? Answer: Continuation: gloober Blue will turn yellow on turn 4. > I find it a little unsettling that GPT-3 is able to do this well. I agree and think that this is definitely the right reaction.
- mbil 4y ago
  Reminds me of a paper that I saw the other day: "On the Paradox of Learning to Reason from Data"[0] [0]: http://starai.cs.ucla.edu/papers/ZhangArxiv22.pdf http://starai.cs.ucla.edu/papers/ZhangArxiv22.pdf
  - foota 4y ago
    This is fantastic, thanks for sharing!
- nikkwong 4y ago
  I know that GPT-3 is not AGI nor modeled properly to ever represent what we would consider to be AGI; but this feels freakishly AGI-ish. This feels like the model is 'reasoning' in a way that I would not expect based on my intuition of how GPT-3 works. As it's been explained to me in lay terms; GPT is great at predicting the next word based on a chunk of previous context. This feels like it's doing much more than that as it's understanding the context and parameters that relate to a question embedded in the text, which seems unrelated to the original instruction of GPT-3 as I understand. Can anyone explain?
  - 4y ago

mjburgess 4y ago

The problem with research conducted in this way is that there's no theory of operation of the model, except the implicit one "it works like how I expect". And then there's no hypothesis test against predictions made based on this theory, ie., no one tries to disconfirm the model works that way. It is usually trivial to disconfirm. ML models almost always work some other way, usually exploiting coincidental statistical information in the training set. This turns the whole show into a superstitious activity, in which we think google answers our "local restaurants" query because it knows where I live (rather than say, using an IP location or one of 100 other methods). If you present a ML model as "capable of x", you're required to disprove that claim as best you can, and present the occasions where it failed.

xyzzy123 4y ago
It might not be rigorous science but it's a cracker of a blog post. It shows a surprising ability of language models, code is provided and plenty of mystery is left for the reader... which can be investigated by almost anyone for tens of dollars.
- mjburgess 4y ago
  But is it an ability? I can't tell. What would a useless "random" algorithm find operating on these graphs as inputs? What would a useful one find? What would that be "useful" for? Is this system much closer to the useless or the useful? We cant tell. Without any explanation, we're left reading this blog post as a magic trick in the literal sense. Consider the rate at which it "uses edges that don't exist" -- this makes it not a graph traversal algorithm. So it is just exploting syntax in the input whose representation has some "graphlike" qualities which can be statistically exploited. Why is that interesting? I dont know. Is this a good system?
  - xyzzy123 4y ago
    I don't know! That's why it's fun. (EDIT: I can safely say this isn't practically useful... but it's interesting in the same way as watching a cat try to fill out a tax return. I'm not surprised the cat does it badly, I'm surprised that it seems to be doing it at all, even if it's really a trick).

taberiand 4y ago

The performance of this model using English descriptions suggests to me the possibility of a model that is built to work using precise mathematical notation performing even better. Are there models that are trained to receive and respond with a purely mathematical description of the problem?

phi0 4y ago
OpenAI has trained a model to solve Math Olympiad-type problems[1], with some initial success, but currently the model isn't good at coming up with proofs involving more than a couple arguments chained together. Still some very impressive work. [1] https://analyticsindiamag.com/openais-neural-theorem-prover-can-solve-math-olympiad-problems/ https://analyticsindiamag.com/openais-neural-theorem-prover-...
lpasselin 4y ago
Check out Google Minerva. https://ai.googleblog.com/2022/06/minerva-solving-quantitative-reasoning.html https://ai.googleblog.com/2022/06/minerva-solving-quantitati...

dr_dshiv 4y ago

Propositions: 1. GPT3 is AGI 2. In combination with people, GPT3 is superintelligence Strong claims. Anyone care to argue for or against?

jmfldn 4y ago
There is zero evidence for the AGI claim or even a weaker intelligence claim. GPT 3 is just a search engine on steroids. I'm regularly suprised by the results but it's honestly just an illusion as far as 'intelligence' goes. This has nothing to do with intelligence, and anyone who thinks otherwise is succumbing to the Eliza effect.
- hackerlight 4y ago
  Why can't the same thing be said of human intelligence? If we knew more about neuroscience it'd probably seem just as trivial. Wetware trained on examples we've seen. Experience encoding neurons. Neurons running on the laws of physics. Data goes in, endogenous processing happens, then outputs. I don't see why it's categorically different. ANNs are just demystified because we made them and know how they work.
  - jmfldn 4y ago
    You've answered your own question. "why can't the same thing be said of human intelligence" "if we knew more about neuroscience..." So that's why we can't say the same thing. Fwiw, my take is that whatever the brain does, its a machine of some sort but possibly not even a turing machine. Many disagree but noone can prove me wrong. Yet at least.
    - pennaMan 4y ago
      So are we intelligence or just a pattern recognition system on steroids?
      - jmfldn 4y ago

bilsbie 4y ago

Where’s the best place to try out one of these language models right now? Doesn’t have to be gpt3.

theblazehen 4y ago
Try signing up for https://beta.openai.com/playground https://beta.openai.com/playground, I got access to GPT-3 in a couple days. For GPT-2 you can try out https://transformer.huggingface.co/ https://transformer.huggingface.co/

jcims 4y ago

Would be interesting to see if it does better if the process was broken down into ’moves’ and the model asked/reminded what its next possible steps are along the way.

jokethrowaway 4y ago

I can't wait for the next version which will surely know how to pass a leetcode interview.

rytill 4y ago

Do other large language models perform well on this task, e.g. EleutherAI models?

ibestvina 4y ago

Great writeup! I made a couple of changes in your code and it seems it gets ~90% of examples correct (inside the first three categories you have) with aroung 65% having either optimal path or correctly saying that there is no path. What I changed was: 1. Used code-davinci-002 (codex) 2. Instead of using an explanation of how these tasks work, I changed your code prompts slightly and instead gave it 3 examples (two with paths, and one to show it what to output when there is no path present). 3. Changed the output so that it has to tell me which edges, instead of which nodes, it is traversing (this helps GPT to avoid using nonexistent edges it seems). Here is an example: https://pastebin.com/D7Hn95VC https://pastebin.com/D7Hn95VC Last "Problem" is the real problem we want to solve of course, everything else is static. I'll post my code when I finish testing :) EDIT: I used Codex not because I think it's better suited for this, but because it's currently free. I don't have enough credit ATM to run 1000 iterations with text-davinci, so I have no idea what difference using Codex made.

tibbar 4y ago
This looks amazing, I’m looking forward to seeing your version!
- ibestvina 4y ago
  Here are the stats :) - Found the optimal path 504 - Found a (non-optimal) path 293 - Correctly reported no path exists 131 - Found a solution when none existed 19 - Ended on wrong node 0 - Used edges that don't exist 53 - Started with the wrong node 0 - Incorrectly reported no path exists 0 - Total: 1000 Here is the code: https://github.com/ibestvina/gpt3-graph-search https://github.com/ibestvina/gpt3-graph-search