7 ms·

When are we going to define what intelligence actually means, so these discussions can start making sense? 5 years ago, it was still common to read things like

by p-e-w 2y ago

When are we going to define what intelligence actually means, so these discussions can start making sense? 5 years ago, it was still common to read things like "AIs will never be as intelligent as mice, let alone humans", today, it's "sure, AIs are as intelligent as some humans, but not as intelligent as the right humans". Noticed how everyone dropped the Turing Test like a hot potato as the gold standard for intelligence, the moment it became apparent that LLMs were about to pass it? Try to find a recent high-profile article invoking the Turing Test. Crickets. The intellectual dishonesty is nauseating. The entire discussion is dominated by smart people who are scared shitless that AI is going to show them just how ordinary they are in the grand scheme of things.

card_zero 2y ago
Do we still have Turing Test competitions? Now that everybody's used to LLMs, and detecting their style, it might be interesting. I'm not sure the LLMs would do very well against humans who were experienced and alert.
- viraptor 2y ago
  "detecting their style" is trivial only because people take the default output and run with it. You can significantly alter it if you want and it would be an obvious thing to do for such test. Right now the default output styles of public LLMs have no reason to try avoiding detection and is not trained to do it.
  - card_zero 2y ago
    I didn't exactly mean superficial style, but properties like creativity, liveliness, and consistency. I hesitate to say that these are things LLMs can't do, because I'll provoke somebody into telling me that they totally can, but weaknesses like this could be ways for human intuition to detect an LLM. It might have to be a long competition, because an unusual set of prompts (or training, even) is likely to initially impress the humans: time should be allowed for novelty to wear off.
    - viraptor 2y ago
      > properties like creativity, liveliness, and consistency I don't believe we train The LLMs for those things specifically. It will be interesting to see if some datasets for this appear. I think we can still make huge improvements by just caring about that more. We've got Samantha though, so I hope we will see those attributes too. https://erichartford.com/meet-samantha https://erichartford.com/meet-samantha
BriggyDwiggs42 2y ago
Wait we say LLMs are as smart as some humans? I must have missed the memo.
- p-e-w 2y ago
  I mean LLMs crush most humans (even many professionals) at coding/legal/medical/etc. exams. Ah, wait. That only counts as evidence of intelligence when humans do it, right?
  - wavemode 2y ago
    Those same humans the AI beats at one task, the human would beat it at 1000 others. So the human is 1000x more intelligent! Ah, wait. That only counts as evidence of intelligence when AIs do it, right?
    - ben_w 2y ago
      It's the other way around. Any one human will only beat an AI at two or three tasks — the first two being our day job and our mother tongue, the third possibly being our hobby — while the AI will out-perform us at thousands of others tasks. Only if you count humanity as a whole do we beat an LLM at everything.
      - wavemode 2y ago
        There are two logical problems with what you've written here. 1. You're counting "our day job" as one task, while counting all individual prompts an AI can answer as each being their own task. This is obviously misleading (and I could just as easily perform the same compression in the opposite direction - chatbots can only do one task: "be a chatbot"). 2. You're not controlling for training. It's already meaningless to compare the intelligence of one entity trained to do a task with another entity that was not trained to do that task. But even ignoring those fallacies, what you've written here is still not true. 99% of things an LLM could supposedly "out-perform" a human at, the human would actually outperform if you provided that human with the same text resources the LLM used to conjure its answer. Regurgitating facts is not evidence of intelligence, and humans can do it easily (and do it without hallucinating, which is key) if you just give them access to the same information. But when you go in the opposite direction, achieving parity is no longer so simple. When LLM's fail to do math, fail to strategize, fail to process rules of abstract games, etc. there is no textbook or lecture or article you can provide to the LLM that magically makes the problem go away. They are fundamental limitations of the AI's capabilities, rather than just a result of not possessing enough information. What many people don't seem to realize is, when merely getting up in the morning and brushing your teeth you are already exercising more intelligence than any AI has ever possessed. (Anyone who has ever worked in robotics or visual processing can attest to that enthusiastically.) So don't even get me started on actual critical thinking.
ben_w 2y ago
> Noticed how everyone dropped the Turing Test like a hot potato as the gold standard for intelligence, the moment it became apparent that LLMs were about to pass it? Try to find a recent high-profile article invoking the Turing Test. Crickets. The intellectual dishonesty is nauseating. Not quite that bad, but I hear you. https://arxiv.org/html/2405.08007v1 https://arxiv.org/html/2405.08007v1

by p-e-w 2y ago

card_zero 2y ago

Do we still have Turing Test competitions? Now that everybody's used to LLMs, and detecting their style, it might be interesting. I'm not sure the LLMs would do very well against humans who were experienced and alert.

viraptor 2y ago
"detecting their style" is trivial only because people take the default output and run with it. You can significantly alter it if you want and it would be an obvious thing to do for such test. Right now the default output styles of public LLMs have no reason to try avoiding detection and is not trained to do it.
- card_zero 2y ago
  I didn't exactly mean superficial style, but properties like creativity, liveliness, and consistency. I hesitate to say that these are things LLMs can't do, because I'll provoke somebody into telling me that they totally can, but weaknesses like this could be ways for human intuition to detect an LLM. It might have to be a long competition, because an unusual set of prompts (or training, even) is likely to initially impress the humans: time should be allowed for novelty to wear off.
  - viraptor 2y ago
    > properties like creativity, liveliness, and consistency I don't believe we train The LLMs for those things specifically. It will be interesting to see if some datasets for this appear. I think we can still make huge improvements by just caring about that more. We've got Samantha though, so I hope we will see those attributes too. https://erichartford.com/meet-samantha https://erichartford.com/meet-samantha

BriggyDwiggs42 2y ago

Wait we say LLMs are as smart as some humans? I must have missed the memo.

p-e-w 2y ago
I mean LLMs crush most humans (even many professionals) at coding/legal/medical/etc. exams. Ah, wait. That only counts as evidence of intelligence when humans do it, right?
- wavemode 2y ago
  Those same humans the AI beats at one task, the human would beat it at 1000 others. So the human is 1000x more intelligent! Ah, wait. That only counts as evidence of intelligence when AIs do it, right?
  - ben_w 2y ago
    It's the other way around. Any one human will only beat an AI at two or three tasks — the first two being our day job and our mother tongue, the third possibly being our hobby — while the AI will out-perform us at thousands of others tasks. Only if you count humanity as a whole do we beat an LLM at everything.
    - wavemode 2y ago
      There are two logical problems with what you've written here. 1. You're counting "our day job" as one task, while counting all individual prompts an AI can answer as each being their own task. This is obviously misleading (and I could just as easily perform the same compression in the opposite direction - chatbots can only do one task: "be a chatbot"). 2. You're not controlling for training. It's already meaningless to compare the intelligence of one entity trained to do a task with another entity that was not trained to do that task. But even ignoring those fallacies, what you've written here is still not true. 99% of things an LLM could supposedly "out-perform" a human at, the human would actually outperform if you provided that human with the same text resources the LLM used to conjure its answer. Regurgitating facts is not evidence of intelligence, and humans can do it easily (and do it without hallucinating, which is key) if you just give them access to the same information. But when you go in the opposite direction, achieving parity is no longer so simple. When LLM's fail to do math, fail to strategize, fail to process rules of abstract games, etc. there is no textbook or lecture or article you can provide to the LLM that magically makes the problem go away. They are fundamental limitations of the AI's capabilities, rather than just a result of not possessing enough information. What many people don't seem to realize is, when merely getting up in the morning and brushing your teeth you are already exercising more intelligence than any AI has ever possessed. (Anyone who has ever worked in robotics or visual processing can attest to that enthusiastically.) So don't even get me started on actual critical thinking.

ben_w 2y ago

> Noticed how everyone dropped the Turing Test like a hot potato as the gold standard for intelligence, the moment it became apparent that LLMs were about to pass it? Try to find a recent high-profile article invoking the Turing Test. Crickets. The intellectual dishonesty is nauseating. Not quite that bad, but I hear you. https://arxiv.org/html/2405.08007v1 https://arxiv.org/html/2405.08007v1