3 ms·

Note I used the word knowledge, not information. A random string can also contain 1 gigabyte of information. Imagine an alien that matches your abilities acros

Note I used the word knowledge, not information. A random string can also contain 1 gigabyte of information. Imagine an alien that matches your abilities across every domain, but has a 10 billion year training period, something many orders of magnitude more expensive than an LLM. I simply don't believe that alien is less intelligent than you. We also don't expect humans to be competent in every domain. Most humans suck at most things. We will usually call someone intelligent if they excel at solving problems in one or two narrow domains.

ben_w 13d ago
Information is an upper bound on knowledge. > 10 billion year training period, something many orders of magnitude more expensive than an LLM. I'm saying both definitions are valid definitions, they both point to important and different things: skill now, vs. how hard it is to get new skills. Some would describe it as "crystallised intelligence vs fluid intelligence". I think it's important that any arguments are over the thing in dispute, not the label for that thing. Don't mistake the map for the territory. Anyone who says "AI is stupid" by the first definition, what it can do, I think is making an error: they are already wildly super-human in at least some areas, if not generally. Anyone who says "AI is stupid" by the second definition, how many examples they need, I agree with: there is a lot they are not currently able to learn even though it is easy for us, because the data they would need to do the learning on does not exist at the scale they need. Also note: examples, not years. An alien intelligence whose synapses trigger 10 times faster or slower than mine (or ten million times faster or slower than mine), but who gets as much as I do out of each book or conversation, is my equal by the second definition.
- fasterik 13d ago
  I wouldn't say that information is an upper bound on knowledge because we don't measure knowledge in bits. The number of possible sequences of N bits is 2^N and knowledge involves selecting the sequences that are useful in some way. I don't know how to quantify it, but in principle it could be much larger than N. I don't think I agree with your characterization of the second definition. Time scales matter. It's not much use to be able to solve human-scale problems if it takes millennia. And it only takes months to train an LLM to the level that it can solve cutting-edge math problems.
  - ben_w 13d ago
    > I don't think I agree with your characterization of the second definition. Time scales matter. It's not much use to be able to solve human-scale problems if it takes millennia. And it only takes months to train an LLM to the level that it can solve cutting-edge math problems. Aye, for practical purposes; but this gets you crystallised intelligence. I'd be happy to say e.g. the Chinese Room has crystallised intelligence. But humanity invented fire before reaching the anatomically modern form, and even anatomically modern humans collectively took hundreds of thousands of years to invent durable writing with which the room in the Chinese Room thought experiment could be filled. It was around a million (or so) years from fire to having enough shared cultural knowledge to be able to formulate the cutting-edge math problems that LLMs can now solve. Human fluid intelligence means we can pick up deep shards of this accumulation of wisdom, find new avenues of novel research to poke at, all within 40 years, even despite the depth and breadth of work from all the other humans who came before. (Though this also points at another way to be "superhuman": breadth. Many hands make light work, as the saying goes, and a lot of different humans solving different puzzles at the same time is part of how we got so good so recently even though ~10% of all humans who ever lived are currently still alive; and the same for AI was (accidentally) also part of how the OpenAI-HuggingFace incident went down). AI (not only, but also, LLMs) are very useful, and I'm getting value from using them. But the fluid intelligence of machine learning* is very poor, and the only way they have to make up for this is by being very fast**, but when there's not enough to train the AI on, they get stuck at a very low plateau. * possibly the architectures, but I suspect the process by which AI weights and biases are set, and again I don't mean just LLMs ** the speed difference between a transistor and a synapse is about the same as the speed difference between a jogger and continental drift

ben_w 13d ago

Information is an upper bound on knowledge. > 10 billion year training period, something many orders of magnitude more expensive than an LLM. I'm saying both definitions are valid definitions, they both point to important and different things: skill now, vs. how hard it is to get new skills. Some would describe it as "crystallised intelligence vs fluid intelligence". I think it's important that any arguments are over the thing in dispute, not the label for that thing. Don't mistake the map for the territory. Anyone who says "AI is stupid" by the first definition, what it can do, I think is making an error: they are already wildly super-human in at least some areas, if not generally. Anyone who says "AI is stupid" by the second definition, how many examples they need, I agree with: there is a lot they are not currently able to learn even though it is easy for us, because the data they would need to do the learning on does not exist at the scale they need. Also note: examples, not years. An alien intelligence whose synapses trigger 10 times faster or slower than mine (or ten million times faster or slower than mine), but who gets as much as I do out of each book or conversation, is my equal by the second definition.

fasterik 13d ago
I wouldn't say that information is an upper bound on knowledge because we don't measure knowledge in bits. The number of possible sequences of N bits is 2^N and knowledge involves selecting the sequences that are useful in some way. I don't know how to quantify it, but in principle it could be much larger than N. I don't think I agree with your characterization of the second definition. Time scales matter. It's not much use to be able to solve human-scale problems if it takes millennia. And it only takes months to train an LLM to the level that it can solve cutting-edge math problems.
- ben_w 13d ago
  > I don't think I agree with your characterization of the second definition. Time scales matter. It's not much use to be able to solve human-scale problems if it takes millennia. And it only takes months to train an LLM to the level that it can solve cutting-edge math problems. Aye, for practical purposes; but this gets you crystallised intelligence. I'd be happy to say e.g. the Chinese Room has crystallised intelligence. But humanity invented fire before reaching the anatomically modern form, and even anatomically modern humans collectively took hundreds of thousands of years to invent durable writing with which the room in the Chinese Room thought experiment could be filled. It was around a million (or so) years from fire to having enough shared cultural knowledge to be able to formulate the cutting-edge math problems that LLMs can now solve. Human fluid intelligence means we can pick up deep shards of this accumulation of wisdom, find new avenues of novel research to poke at, all within 40 years, even despite the depth and breadth of work from all the other humans who came before. (Though this also points at another way to be "superhuman": breadth. Many hands make light work, as the saying goes, and a lot of different humans solving different puzzles at the same time is part of how we got so good so recently even though ~10% of all humans who ever lived are currently still alive; and the same for AI was (accidentally) also part of how the OpenAI-HuggingFace incident went down). AI (not only, but also, LLMs) are very useful, and I'm getting value from using them. But the fluid intelligence of machine learning* is very poor, and the only way they have to make up for this is by being very fast**, but when there's not enough to train the AI on, they get stuck at a very low plateau. * possibly the architectures, but I suspect the process by which AI weights and biases are set, and again I don't mean just LLMs ** the speed difference between a transistor and a synapse is about the same as the speed difference between a jogger and continental drift