Why can't we find a better AI architecture than transformers?
>>108231218we've found plenty, they just don't scale.
>>108231247Why come the can't scale
>>108231322they just.. le dont ok chud?
>>108231218research takes time, tech companies prefer to throw billions of hardware to it instead of doing proper researchlike that is the diferent with before, that since the high speed interconections and the bandwith of new hardware, we can just throw more hardware to the problem in the most retarded way
>>108231218because research hasn't been the goal for years now
>>108232053people miss this, we are barely researching at all
>>108231322probably the same reason the transformer architecture wasn't viable until recently.
>>108232053They're just going to milk one architecture leap forever then
>>108234805We are actually researching more than ever but making fewer discoveries and breakthroughs. Eventually, there won't be any progression in science and technology due to human limits.
