@BGatesIsaPyscho Western civilization went through the period of enlightenment in the 18th century when we discovered freedom from religious oppression, and entered into secular society. Religion is a private matter and should be kept separate from education and politics. Don’t go back!!
@fchollet Are you suggesting the brain evolved to discover meta-rules to adapt its own architecture? Do you have any argument for this claim? Or evidence? For example the meta-rules that resulted in the language faculty?
@MelMitchell1 Are there any principles to discover? And if there are, what use are they? Do they help build better models? Or just reassure us that LLMs don’t think like us. We kind of know that already though.
@MelMitchell1 As a cognitive scientist I couldn’t agree more with these principles of discovery. But I wonder why we want to discover how AI models “think”. Natural Sciences uncover how the world works. AI is an engineering artifact and their “thinking” changes from model to model.
@DeryaTR_ This is not a rebuttal. It is mere terminological imperialism. Yann at least gave an argument of sorts for his position, Demis just throws around some words.
@Kon__K That does not contradict what Hastie said. Ahmed is a migrant who more than lived up to Australian values. So how is that an argument against Hastie?
@David_Tomasian@Rapahelz@fchollet More precisely, they regurgitate what they have already seen, and by definition these discoveries would be things that have never been written about or thought about.
@David_Tomasian@Rapahelz@fchollet But the claim is "... to create an AGI that is more capable than humans of grasping the strangeness of the universe." But even if we could create such an AGI, we could never understand what it has discovered. But since a NN can't grasp "anything" this is unlikely.
@David_Tomasian@Rapahelz@fchollet It helps us because it exposes the arrogance of the view being presented. Chomsky has also pointed out that as biological creatures we are bound by the same rules as other species. Rats can’t understand complex numbers. We can’t understand the third kind of causation
@fchollet That first paragraph is gibberish. What is “sufficiently general”? Science doesn’t model anything; it’s a method. If I was to grant that there is an argument at all then it is circular.
@QuentinDempster@sussanley@AlboMP They didn’t say incited. They said it creates an environment where radicals can operate. It’s like mixing the prehistoric soup where life first emerged.
@fchollet LeCun shows insight when he says we can solve any problem we can imagine. Which is very different from “solvable problem”. Chomsky made this point many years ago. He said rats can’t imagine complex numbers, and we basically imagine two forms of causality. What if there are more?