4 ms·

Nope, there's no reasoning. It's just generating the text that best matches its training data. They admit that themselves, which makes the statement "bard is ge

Nope, there's no reasoning. It's just generating the text that best matches its training data. They admit that themselves, which makes the statement "bard is getting better at reasoning" even more irritating: > Large language models (LLMs) are like prediction engines — when given a prompt, they generate a response by predicting what words are likely to come next

wilg 3y ago
Bard can be an AI product that includes an LLM among other things and that is fine and good.
ajuc 3y ago
> Nope, there's no reasoning. It's just generating the text that best matches its training data. There's no contradiction. You have to reason to predict the text well in many cases.
- jerf 3y ago
  Probably the best answer is, "The concept in your head labelled by 'reasoning' doesn't apply, but neither does the one you associate with 'unreasoning'." It isn't doing classical reasoning per se, but neither does it match an unreasoning brute process. In general, you should get used to this. Probably every AI from this point on out until they simply exceed us entirely and we can't mentally model them at all are going to be neither quite what we consider "human reasoning", but that doesn't mean they are "unreasoning" either. We'll be able to see certain flaws, but then again they will increasingly be able to see our flaws too.
  - ajuc 3y ago
    > It isn't doing classical reasoning per se, but neither does it match an unreasoning brute process. What we call "classical reasoning" is also running on neural network trained by predicting data. I think the distinction is only there because we mix levels of abstraction - when talking about AI we focus on the lowest levels, when talking about people we focus on the higher levels. Try asking LLM to explain its reasoning and after dismissing the "I'm just AI" disclaimers it will usually generate logical chain that could just as well be generated by a human. Even if you think it's not "really" reasoning - just pretending - pretending means it's running a virtual machine and that virtual machine is reasoning. Same difference.
HarHarVeryFunny 3y ago
> Nope, there's no reasoning. It's just generating the text that best matches its training data. That's like saying that when you answer questions on an exam, you're just generating the text that best matches your training data... Both statements are correct, but only if you understand what "generating" and "matches" mean. Generating doesn't (always) mean copying, and matches doesn't (always) mean exactly the same. In the more general case you're drawing a kind of analogy between what you were taught and the new problem you are answering. You should google "Induction heads" which is one of the mechanisms that researchers believe Transformers are using to perform in-context learning. In the general case this is an analogical A'B' => AB type of "prediction".

wilg 3y ago

Bard can be an AI product that includes an LLM among other things and that is fine and good.

ajuc 3y ago

> Nope, there's no reasoning. It's just generating the text that best matches its training data. There's no contradiction. You have to reason to predict the text well in many cases.

jerf 3y ago
Probably the best answer is, "The concept in your head labelled by 'reasoning' doesn't apply, but neither does the one you associate with 'unreasoning'." It isn't doing classical reasoning per se, but neither does it match an unreasoning brute process. In general, you should get used to this. Probably every AI from this point on out until they simply exceed us entirely and we can't mentally model them at all are going to be neither quite what we consider "human reasoning", but that doesn't mean they are "unreasoning" either. We'll be able to see certain flaws, but then again they will increasingly be able to see our flaws too.
- ajuc 3y ago
  > It isn't doing classical reasoning per se, but neither does it match an unreasoning brute process. What we call "classical reasoning" is also running on neural network trained by predicting data. I think the distinction is only there because we mix levels of abstraction - when talking about AI we focus on the lowest levels, when talking about people we focus on the higher levels. Try asking LLM to explain its reasoning and after dismissing the "I'm just AI" disclaimers it will usually generate logical chain that could just as well be generated by a human. Even if you think it's not "really" reasoning - just pretending - pretending means it's running a virtual machine and that virtual machine is reasoning. Same difference.

HarHarVeryFunny 3y ago

> Nope, there's no reasoning. It's just generating the text that best matches its training data. That's like saying that when you answer questions on an exam, you're just generating the text that best matches your training data... Both statements are correct, but only if you understand what "generating" and "matches" mean. Generating doesn't (always) mean copying, and matches doesn't (always) mean exactly the same. In the more general case you're drawing a kind of analogy between what you were taught and the new problem you are answering. You should google "Induction heads" which is one of the mechanisms that researchers believe Transformers are using to perform in-context learning. In the general case this is an analogical A'B' => AB type of "prediction".