4 ms·

LLMs are notoriously terrible at multiplying large numbers: https://claude.ai/share/538f7dca-1c4e-4b51-b887-8eaaf7e6c7d3 https://claude.ai/share/538f7dca-1c4e-4

by zamalek 7mo ago

LLMs are notoriously terrible at multiplying large numbers: https://claude.ai/share/538f7dca-1c4e-4b51-b887-8eaaf7e6c7d3 https://claude.ai/share/538f7dca-1c4e-4b51-b887-8eaaf7e6c7d3 > Let me calculate that. 729,278,429 × 2,969,842,939 = 2,165,878,555,365,498,631 Real answer is: https://www.wolframalpha.com/input?i=729278429*2969842939 https://www.wolframalpha.com/input?i=729278429*2969842939 > 2 165 842 392 930 662 831 Your example seems short enough to not pose a problem.

peacebeard 7mo ago
This doesn’t address the author’s point about novelty at all. You don’t need 100% accuracy to have the capability to solve novel problems.
- zamalek 7mo ago
  It does address the GP comment about math.
luma 7mo ago
Modern LLMs, just like everyone reading this, will instead reach for a calculator to perform such tasks. I can't do that in my head either, but a python script can so that's what any tool-using LLM will (and should) do.
- zamalek 7mo ago
  This is special pleading. Long multiplication is a trivial form of reasoning that is taught at elementary level. Furthermore, the LLM isn't doing things "in its head" - the headline feature of GPT LLMs is attention across all previous tokens, all of its "thoughts" are on paper. That was Opus with extended reasoning, it had all the opportunity to get it right, but didn't. There are people who can quickly multiply such numbers in their head (I am not one of them). LLMs don't reason.
  - markatto 7mo ago
    I tried this with Claude - it has to be explicitly instructed to not make an external tool call, and it can get the right answer if asked to show its work long-form.
  - esafak 7mo ago
    Mathematics is not the only kind of reasoning, so your conclusion is false. The human brain also has compartments for different types of activities. Why shouldn't an AI be able to use tools to augment its intelligence?
    - zamalek 7mo ago
      I used the mathematics example only because the GP did. There are many other examples of non-reasoning, including some papers (as recent as Feb).
      - azakai 7mo ago
        There are many examples of current limitations, but do you see a reason to think they are fundamental limitations? (I'm not saying they aren't, I'm curious what the evidence is for that.)
        zamalek 7mo ago
        It's because of how transformers work, especially the fact that the output layer is a bunch of weights which we quite literally do a weighted random choice from. My hunch is that diffusion models would have a higher chance of doing real reasoning - or something like a latent space for reasoning. Thinking that LLMs are intelligent arises from an incomplete understanding of how they work or, alternatively, having shareholders to keep happy.
  - CamperBob2 7mo ago
    Furthermore, the LLM isn't doing things "in its head" - the headline feature of GPT LLMs is attention across all previous tokens, all of its "thoughts" are on paper LOL, talk about special pleading. Whatever it takes to reshape the argument into one you can win, I guess... LLMs don't reason. Let's see you do that multiplication in your head. Then, when you fail, we'll conclude you don't reason. Sound fair?
- jibal 7mo ago
  LLMs don't use tools. Systems that contain LLMs are programmed to use tools under certain circumstances.
  - dirasieb 7mo ago
    you’re just abstracting it away into this new “systems” definition when someone says LLMs today they obviously mean software that does more than just text, if you want to be extra pedantic you can even say LLMs by themselves can’t even geenrate text since they are just model files if you don’t add them to a “system” that makes use of that model files, doh
    - jibal 7mo ago
      > when someone says LLMs today they obviously mean ... LLMs, if the someone is me or others who understand why it's important to be precise. And in this context, the distinction between LLM and AI mattered--not pedantic at all. I won't respond further ... over and out.
lern_too_spel 7mo ago
This hasn't been true for a while now. I asked Gemini 3 Thinking to compute the multiplication "by hand." It showed its work and checked its answer by casting out nines and then by asking Python. Sonnet 4.6 with Extended Thinking on also computed it correctly with the same prompt.
throwawayk7h 7mo ago
I thought it might do better if I asked it to do long-form multiplication specifically rather than trying to vomit out an answer without any intermediate tokens. But surprisingly, I found it doesn't do much better.
- zamalek 7mo ago
  Other comments indicate that asking it to do long multiplication does work, but the varying results makes sense: LLMs are probabilistic, you probably rolled an unlikely result.
  - lern_too_spel 7mo ago
    Specifically, you need to use a reasoning model. Applying more test time compute is analogous to Kahneman's System 2 thinking, while directly taking the first output of an LLM is analogous to System 1. This is true for solving difficult novel problems as well, with the addition of tools that an agent can use to research the problem autonomously.

by zamalek 7mo ago

peacebeard 7mo ago

This doesn’t address the author’s point about novelty at all. You don’t need 100% accuracy to have the capability to solve novel problems.

zamalek 7mo ago
It does address the GP comment about math.

luma 7mo ago

Modern LLMs, just like everyone reading this, will instead reach for a calculator to perform such tasks. I can't do that in my head either, but a python script can so that's what any tool-using LLM will (and should) do.

zamalek 7mo ago
This is special pleading. Long multiplication is a trivial form of reasoning that is taught at elementary level. Furthermore, the LLM isn't doing things "in its head" - the headline feature of GPT LLMs is attention across all previous tokens, all of its "thoughts" are on paper. That was Opus with extended reasoning, it had all the opportunity to get it right, but didn't. There are people who can quickly multiply such numbers in their head (I am not one of them). LLMs don't reason.
- markatto 7mo ago
  I tried this with Claude - it has to be explicitly instructed to not make an external tool call, and it can get the right answer if asked to show its work long-form.
- esafak 7mo ago
  Mathematics is not the only kind of reasoning, so your conclusion is false. The human brain also has compartments for different types of activities. Why shouldn't an AI be able to use tools to augment its intelligence?
  - zamalek 7mo ago
    I used the mathematics example only because the GP did. There are many other examples of non-reasoning, including some papers (as recent as Feb).
    - azakai 7mo ago
      There are many examples of current limitations, but do you see a reason to think they are fundamental limitations? (I'm not saying they aren't, I'm curious what the evidence is for that.)
      - zamalek 7mo ago
        It's because of how transformers work, especially the fact that the output layer is a bunch of weights which we quite literally do a weighted random choice from. My hunch is that diffusion models would have a higher chance of doing real reasoning - or something like a latent space for reasoning. Thinking that LLMs are intelligent arises from an incomplete understanding of how they work or, alternatively, having shareholders to keep happy.
- CamperBob2 7mo ago
  Furthermore, the LLM isn't doing things "in its head" - the headline feature of GPT LLMs is attention across all previous tokens, all of its "thoughts" are on paper LOL, talk about special pleading. Whatever it takes to reshape the argument into one you can win, I guess... LLMs don't reason. Let's see you do that multiplication in your head. Then, when you fail, we'll conclude you don't reason. Sound fair?
jibal 7mo ago
LLMs don't use tools. Systems that contain LLMs are programmed to use tools under certain circumstances.
- dirasieb 7mo ago
  you’re just abstracting it away into this new “systems” definition when someone says LLMs today they obviously mean software that does more than just text, if you want to be extra pedantic you can even say LLMs by themselves can’t even geenrate text since they are just model files if you don’t add them to a “system” that makes use of that model files, doh
  - jibal 7mo ago
    > when someone says LLMs today they obviously mean ... LLMs, if the someone is me or others who understand why it's important to be precise. And in this context, the distinction between LLM and AI mattered--not pedantic at all. I won't respond further ... over and out.

lern_too_spel 7mo ago

This hasn't been true for a while now. I asked Gemini 3 Thinking to compute the multiplication "by hand." It showed its work and checked its answer by casting out nines and then by asking Python. Sonnet 4.6 with Extended Thinking on also computed it correctly with the same prompt.

throwawayk7h 7mo ago

I thought it might do better if I asked it to do long-form multiplication specifically rather than trying to vomit out an answer without any intermediate tokens. But surprisingly, I found it doesn't do much better.

zamalek 7mo ago
Other comments indicate that asking it to do long multiplication does work, but the varying results makes sense: LLMs are probabilistic, you probably rolled an unlikely result.
- lern_too_spel 7mo ago
  Specifically, you need to use a reasoning model. Applying more test time compute is analogous to Kahneman's System 2 thinking, while directly taking the first output of an LLM is analogous to System 1. This is true for solving difficult novel problems as well, with the addition of tools that an agent can use to research the problem autonomously.