Gary Marcus should be disregarded because he’s emotionally invested in The Bitter Lesson being wrong. He really wants LLMs to not be as good as they already are. He’ll find some interesting research about “here’s a limitation that we found” and turn that into “LLMS BTFO IT’S SO OVER”.
The research is interesting for helping improve LLMs, but that’s the extent of it. I would not be worried about the limitations the paper found for a number of reasons:
There doesn’t seem to be any reason to believe that there’s a ceiling on scaling up
LLM’s reasoning abilities improve with scale (notice that the example they use for kiwis they included the answers from o1-mini and llama3-8B, which are much smaller models with much more limited capabilities. GPT-4o got the problem correct when I tested it, without any special prompting techniques or anything)
Techniques such as RAG and Chain of Thought help immensely on many problems
Basic prompting techniques help, like “Make sure you evaluate the question to ignore extraneous information, and make sure it’s not a trick question”
LLMs are smart enough to use tools. They can go “Hey, this looks like a math problem, I’ll use a calculator”, just like a human would
There’s a lot of research happening very quickly here. For example, LLMs improve at math when you use a different tokenization method, because it changes how the model “sees” the problem
Until we hit a wall and really can’t find a way around it for several years, this sort of research falls into the “huh, interesting” territory for anybody that isn’t a researcher.
Actually we do know that there are diminishing returns from scaling already. Furthermore, I would argue that there are inherent limits in simply using correlations in text as the basis for the model. Human reasoning isn’t primarily based on language, we create an internal model of the world that acts as a shared context. The language is rooted in that model and that’s what allows us to communicate effectively and understand the actual meaning behind words. Skipping that step leads to the problems we’re seeing with LLMs.
That said, I agree they are a tool, and they obviously have uses. I just think that they’re going to be a part of a bigger tool set going forward. Right now there’s an incredible amount of hype associated with LLMs. Once the hype settles we’ll know what use cases are most appropriate for them.
The whole “it’s just autocomplete” is just a comforting mantra. A sufficiently advanced autocomplete is indistinguishable from intelligence. LLMs provably have a world model, just like humans do. They build that model by experiencing the universe via the medium of human-generated text, which is much more limited than human sensory input, but has allowed for some very surprising behavior already.
We’re not seeing diminishing returns yet, and in fact we’re going to see some interesting stuff happen as we start hooking up sensors and cameras as direct input, instead of these models building their world model indirectly through purely text. Let’s see what happens in 5 years or so before saying that there’s any diminishing returns.
Gary Marcus should be disregarded because he’s emotionally invested in The Bitter Lesson being wrong. He really wants LLMs to not be as good as they already are. He’ll find some interesting research about “here’s a limitation that we found” and turn that into “LLMS BTFO IT’S SO OVER”.
The research is interesting for helping improve LLMs, but that’s the extent of it. I would not be worried about the limitations the paper found for a number of reasons:
o1-miniandllama3-8B, which are much smaller models with much more limited capabilities. GPT-4o got the problem correct when I tested it, without any special prompting techniques or anything)Until we hit a wall and really can’t find a way around it for several years, this sort of research falls into the “huh, interesting” territory for anybody that isn’t a researcher.
Actually we do know that there are diminishing returns from scaling already. Furthermore, I would argue that there are inherent limits in simply using correlations in text as the basis for the model. Human reasoning isn’t primarily based on language, we create an internal model of the world that acts as a shared context. The language is rooted in that model and that’s what allows us to communicate effectively and understand the actual meaning behind words. Skipping that step leads to the problems we’re seeing with LLMs.
That said, I agree they are a tool, and they obviously have uses. I just think that they’re going to be a part of a bigger tool set going forward. Right now there’s an incredible amount of hype associated with LLMs. Once the hype settles we’ll know what use cases are most appropriate for them.
The whole “it’s just autocomplete” is just a comforting mantra. A sufficiently advanced autocomplete is indistinguishable from intelligence. LLMs provably have a world model, just like humans do. They build that model by experiencing the universe via the medium of human-generated text, which is much more limited than human sensory input, but has allowed for some very surprising behavior already.
We’re not seeing diminishing returns yet, and in fact we’re going to see some interesting stuff happen as we start hooking up sensors and cameras as direct input, instead of these models building their world model indirectly through purely text. Let’s see what happens in 5 years or so before saying that there’s any diminishing returns.