5 ms·
I think it's reductivism to assume that neural networks cannot emergently support/implement a non-stochastic computational model capable of explicit logical rea
by calf 3y ago
I think it's reductivism to assume that neural networks cannot emergently support/implement a non-stochastic computational model capable of explicit logical reasoning.
We already have an instance of emergent logic. Animals engage in logical reasoning. Corollary, humans and toddlers are not merely super-autocompletes or stochastic parrots.
It has nothing to do with "sensory embodiment" and/or "personal agency" arguments like in the article. Nor the clever solipsism and reductivism of "my mind is just random statistics of neural firing". It's about finding out what the models of computation actually are.
- PaulDavisThe1st 3y agoImagine that for some reason and by some unknown agency, somebody in 1068 could build something that was functionally equivalent to one of the contemporary LLMs of today (it would likely be mechanical and thus slower, but let's just ignore speed because that's mostly a substrate artifact). OK, time to train up our LLM. Oops. Printing not yet invented. Books are scarce, and frequently not copied. We have almost no input data to give to our analog LLM. Will it be of any use whatsoever? I think we can say with a high degree of confidence that it will not. However, contrast a somewhat typical human from the same era. Will they suffer from any fundamental limitation compared to someone alive today? Will there be skills they cannot acquire? Languages they cannot learn? It's not quite so easy to be completely confident about the answer to this, but I think there's a strong case for "no". The data dependence of LLMs compared to human (or ravens) makes a compelling argument that they are not operating in the same way (or, if they are, they have somehow reduced the data dependency in an incredibly dramatic way, such that it may still be reasonable to differentiate the mechanisms).
- gnaritas99 3y agoHumans come on pre-trained hardware. An LLM embodied with senses would have all the data it ever needs starting from a similar pre-training.
- intelkishan 3y agoA common counterargument is that organisms already have information stored in their DNA, which in turn has a noticeable impact on their behaviour. Plus the fact that they receive a large amount of sensory data which is often discounted in such debates.
- cozzyd 3y agoGreat, instead of using a corpus, just use videos from toddlers walking around with a go pro as training data for your AI model and see how far that gets you.
- calf 3y agoWell I think toddlers get both offline training from millions of years of evolution (idea being that evolution itself is nature's very own training algorithm, a computational process that creates genetic information), and online training in the form of nurture and environmental interaction -- and of course AI professors are saying the way to go is to research was to make online training possible for models. But the crux is that both kinds of "training" - almost a loaded term by now - are open areas of scientific research.
- PaulDavisThe1st 3y agoOK, so LLM's (a) have no pre-wiring (b) have no sensory connection but they can still spit out remarkably good language, and thus they must be a good model for humans and animals?
- calf 3y agoI don't find that a compelling argument :) If one computational model accepts only assembly code, and another model accepts only C++, it tells us nothing about what is going on inside their black boxes. They could be doing computationally fundamentally equivalent things, or totally different things. We don't know till we look. My point was really about neural networks in general. The argument is that neural networks include both humans and synthetic ones. The myopic focus on LLMs (because of mainstream media coverage) as they exist now is what limits such arguments.