Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

The Bitter Lesson (2019) [pdf]

cs.utexas.edu

Post reply on HN
Print view View original

21–30 of 62 posts

Re: The Bitter Lesson (2019) [pdf]

#21

Post by michalsustr »

currymj wrote: post #20
isaacfrond wrote: post #7

I think the Go example is facetious. For the longest time, Go simply was not amenable to search. It is only because of breakthrough in theory that Go became amendable to search. And, it is hardly a straightforward search mind you. Two self learning networks are involved. A first to guide the expansion of a Monte Carlo based tree search, a second to evaluate the nodes in the tree. You can hardly blame a researcher in…

i don't know. it's mysterious that nobody even came up with the algorithms for decades, which often do still work well on toy examples. MCTS is the kind of thing people definitely could have thought of at any point since 1950; and neural network value functions for board games had already succeeded with TD-GAMMON. this is even more striking in poker, where counterfactual regret minimization wasn't invented until 2007…

Can you please give a link to the soviet paper? Thank you very much :)

Re: The Bitter Lesson (2019) [pdf]

#22

Post by niemandhier »

psychoslave wrote: post #9
niemandhier wrote: post #8

I think the Bitter Lessons do not consider two important points with regards to neural nets : 1. Not all nets work for all problems, those that work tend to have the right inductive biases. We discovered the architectures partially by trial an error, nevertheless they work because of encoded prior information. 2. Data and computation are bounded. GPT4 was basically trained on all text, further advancements probably n…

What do you mean with more insights? If they are not represented as some data, how could they ever influence any computational model?

By building the model in such a way that assumptions about the data are in the model structure.

ConvNets hardcode translation equivariance, more general convolutions can hardcode more general equivariant structures.

The whole field of geometric deep learning is about constructing nets based on insights about the structure of the data manifold, and many successful nets turn out to do this.

Re: The Bitter Lesson (2019) [pdf]

#23

Post by nologic01 »

Xcelerate wrote: post #12

> The second general point to be learned from the bitter lesson is that the actual contents of minds are tremendously, irredeemably complex; we should stop trying to find simple ways to think about the contents of minds, such as simple ways to think about space, objects, multiple agents, or symmetries. I agree and think “ML explainability” efforts are doomed to fail as ML becomes increasingly more effective. There is…

This (popular) opinion makes no sense. How are you going to improve something you dont understand? Throwing ramdom software pieces to see what sticks? The recent progress was possible because people understood intuitively enough of the limitations of earlier models to think and invent a remedy.

Re: The Bitter Lesson (2019) [pdf]

#24

Post by randomlogic »

nologic01 wrote: post #23
Xcelerate wrote: post #12

> The second general point to be learned from the bitter lesson is that the actual contents of minds are tremendously, irredeemably complex; we should stop trying to find simple ways to think about the contents of minds, such as simple ways to think about space, objects, multiple agents, or symmetries. I agree and think “ML explainability” efforts are doomed to fail as ML becomes increasingly more effective. There is…

This (popular) opinion makes no sense. How are you going to improve something you dont understand? Throwing ramdom software pieces to see what sticks? The recent progress was possible because people understood intuitively enough of the limitations of earlier models to think and invent a remedy.

(Reads:) "Did you know that time-traveling was invented, cos time was seen as a threat to life ?"

P-:

Re: The Bitter Lesson (2019) [pdf]

#25

Post by Tuna-Fish »

nologic01 wrote: post #23
Xcelerate wrote: post #12

> The second general point to be learned from the bitter lesson is that the actual contents of minds are tremendously, irredeemably complex; we should stop trying to find simple ways to think about the contents of minds, such as simple ways to think about space, objects, multiple agents, or symmetries. I agree and think “ML explainability” efforts are doomed to fail as ML becomes increasingly more effective. There is…

This (popular) opinion makes no sense. How are you going to improve something you dont understand? Throwing ramdom software pieces to see what sticks? The recent progress was possible because people understood intuitively enough of the limitations of earlier models to think and invent a remedy.

Not that I necessarily agree with the above poster, but this:

> How are you going to improve something you dont understand?

Is just nonsense. Evolution understands nothing, yet produced a mind. Closer to us, the early people who produced all the crops that led to the shift to agriculture, and it's later improvements, absolutely did not understand how any of it worked.

Evaluation and selection are sufficient to improve things. Understanding is useful, but optional.

Re: The Bitter Lesson (2019) [pdf]

#26

Post by currymj »

michalsustr wrote: post #21
currymj wrote: post #20

Earlier quoted context omitted.

i don't know. it's mysterious that nobody even came up with the algorithms for decades, which often do still work well on toy examples. MCTS is the kind of thing people definitely could have thought of at any point since 1950; and neural network value functions for board games had already succeeded with TD-GAMMON. this is even more striking in poker, where counterfactual regret minimization wasn't invented until 2007…

Can you please give a link to the soviet paper? Thank you very much :)

The title is "Reduction of a game with complete memory to a matrix game" by I. V. Romanovsky, Doklady Akademii Nauk SSSR 1962. Unfortunately I don't have access to the article.

Re: The Bitter Lesson (2019) [pdf]

#27

Post by layer8 »

Xcelerate wrote: post #12

> The second general point to be learned from the bitter lesson is that the actual contents of minds are tremendously, irredeemably complex; we should stop trying to find simple ways to think about the contents of minds, such as simple ways to think about space, objects, multiple agents, or symmetries. I agree and think “ML explainability” efforts are doomed to fail as ML becomes increasingly more effective. There is…

There is also no a-priori reason why we shouldn’t be able to understand how the higher-level behavior emerges. And without such an understanding, trying to improve or control the behavior is like poking around in the dark.

Re: The Bitter Lesson (2019) [pdf]

#30

Post by bonoboTP »

reacweb wrote: post #13
isaacfrond wrote: post #7

I think the Go example is facetious. For the longest time, Go simply was not amenable to search. It is only because of breakthrough in theory that Go became amendable to search. And, it is hardly a straightforward search mind you. Two self learning networks are involved. A first to guide the expansion of a Monte Carlo based tree search, a second to evaluate the nodes in the tree. You can hardly blame a researcher in…

Go has been described as the drosophila of AI. Once a program will know how to play go, we'll be able to make more general artificial intelligences. This is exactly what has happened with alphago and chatgpt. We're not repeating the same thing that happened with chess, but we are past the singularity. Recent advances in AI are lighting up the very nature of intelligence. Our memory and our brain are a machine for pre…

There's no real knowledge transfer from Alphago to Chatgpt. They share some underlying techniques, but they sit on different branches of a tree, one is not built on top of the other.
Post reply on HN