Interpretability researcher Nick Cammarata argues outsmarting humans will require millions of GPUs, framing the compute scale as a reflection of human intelligence
Creator bayes cautioned that hardware scaling limits may not last.
Positive users admire biological nanotech and human knowledge achievements when discussing the millions of GPUs needed for AI to outsmart humans, while others caution against premature optimism.
No Digg Deeper questions have been answered for this story yet.
Most Activity
@nickcammarata Kinda, but the GPUs are making more GPUs soon so not sure if those count
@nickcammarata you really have to appreciate the beauty of biologically derived nanotech
@nickcammarata if we were smarter maybe we could get it done with fewer gpus
@nickcammarata @leothecurious Don’t count the chickens just yet
I think we should have pride as a species that it’s going to take at least millions of gpus to outsmart us. that feels like an impressive amount
@nickcammarata Also, knowledge. We codified knowledge (invented written language). That alone is ridiculous. If that let to machines that recursively find all the words for all possible truth, dope.
@shakermanjonas @nickcammarata i was just gonna say that lol. the more compute it takes to surpass us, the less elegant our solution to intelligence is.