Yann LeCun, Chomsky & Marcus will keep gaslighting on ChatGPT right up until the AI Singularity

They’ve truly lost the plot, and deep down, we know why. ChatGPT has demonstrably solved NLU (natural language understanding) from left-field and largely disproved a subset of 50 years of linguistics and brain psychology theories.

(And I’ve got skin in this game. Our startup lost out too. Our symbolic AI NLU stack is a shadow of its former significance, ever since GPT-1. But after the initial paradigm shock, we’ve been cheering LLMs on, and using them. GPT’s emergent cognition is incredible to say the least).

These three researchers don’t even agree with each other — which is fine — but the only thing I think they are right about is that we need to be careful with AI.

The rest is . . meh and patently deluded . . at best. And unbalanced and wrong at worst.

Recent statements

GPT-4 is arguably 3/4s of the way to AGI (artificial general intelligence) and yet LeCun, one of deep learning’s undeniable heroes, and the only true AI guy amongst the three, here in April 2023, said yesterday that:

Absurd. ChatGPT dog-like? LOL. (I am flippant here only because it’s become abundantly clear GPT is virtually superhuman let alone dog-like, see my posts linked…