at a very basic level neural networks aren’t complicated
you basically just multiply and add lol
the issue is that you do those operations a metric fuck ton of times, and each individual operation has no discernible meaning
arete i’ve been paying attention to the Top AI Researchers at Google and OpenAI and such
they are not actually trying to understand how the machines work
hear me out. use ai to understand ai
all they’re trying to do is trying to make them bigger, not to actually make any meaningful advancements in the field
1 Like
unironically this is basically the only viable path
1 Like
real
I have a feeling that’s an idea already used
Yeah
I have a feeling you get no bitches
Who
yes that’s already been tried it doesn’t work because the machine you’re using to analyse the neural net is just as useless as the original neural net
1 Like
Thank goodness
1 Like
sigmund freud aka john projection
1 Like
analysing our machine models with other machine models would be good if those models were actually good at what they do. but they are profoundly shit at what they do
1 Like
i have a class which is basically just sit there and observe The Horrors. last class we sat there replicating our voices using ai. I do not want to be here
1 Like
his entire shtick is that he thought everyone was as obsessed with the birds and the bees as he was
I hate that people call it artificial intelligence
There is nothing about it that is intelligent
i feel it helps at this point to remind people that sigmund freud and all of his patients were addicted to cocaine
literally the first result from Google for “OpenAI interpretability papers” is a paper where they. are trying to understand LLM neuron activation. this is not hard to find??
like I can’t believe you’re making me defend fucking OpenAI but there are definitely people working on this
1 Like
Artificial, the macchine