The company has done nothing but raise money off of tech that can be mathematically described as a magic trick. Its a literal scam.
by Grimblewald
0 subcomment
Can we stop already? we've demonstrated sufficiently that all known architectures lack the required indicative bias for what the openAI charlatans are Hucking. Sure if you scale a dense feed forward neural network enough it can do vision just fine, but the lack of inductive bias means the model becomes so inefficient that you will _never_ be able to extract any net gain from it.
So it is with transformers. Fantastic ontological cartographers, but not cognitive / logic engines like promised.
by thenaturalist
0 subcomment
"Just another 100 billion and then we'll have [AGI | replaced all software engineers | cured cancer | INSERT LOFTY IMAGINATIVE HYPERBOLE].
Hacker News