The actual meat of the paper is "we got some very slow distributed inferencing working under laboratory conditions with access to homogenous equipment that was bought specifically for this purpose and was already under our control". We've had self-replicating worms since the 1980s at least. This is all very basic stuff. There's absolutely no reason for deploying a distributed LLM as part of this setup.
Search
5 results for “spzb”
-
I mean, look at this shit. “We reckon these steps are necessary for this fantasy to be true. Most of them are difficult so we just hand wave those away and get on with something that’s only tangentially related to the headline “
-
Anthropic are just taking the piss now #Anthropic #aidoomerism #ipo
-
Day job is rolling out an “official” AI platform for internal use. Seems to involve them running training sessions every day for the next few weeks. Can’t help but laugh at the irony of having to train people to use a supposedly intelligent system. If it’s that fucking clever why do I need to know how to prompt it? Why can’t it work it out itself?
(The above is ironic and rhetorical. I already fully understand why it needs some arcane incantation to make it spit out plausible-sounding slop)
-
Groundbreaking research from the Institute for the Blindingly Obvious