Mnet_User

Internet user

[Mnet_User @mnet_user]

@LamBoo35012233 @_albertorod_ @BostonDynamics You're right... We should invent a portable electricity supply! Probably huge money in that

[Mnet_User @mnet_user]

@jon_barron Just citing an LLM isn't helpful... Who cares? Like citing Google search. Authors could add a section to the appendix stating the models and workflows they used. Same as it'd be helpful to also list failed experiments and other dead ends. These actually help other researchers

[Mnet_User @mnet_user]

@deeperflows @ggerganov @NVIDIAAIDev How about opening the links that were added by OP? It is listed very clearly.

[Mnet_User @mnet_user]

@VicVijayakumar Newer models are better at completely re-writing old model slop, instead of fixing old slop. Maybe even by design... Does it burn more tokens?

[Mnet_User @mnet_user]

@felixius_F @RoyalFamily @ylecun He already got this award in the 90s

[Mnet_User @mnet_user]

@predict_addict Updated: Would you agree that forecasting* academia is responsible for many of the explanations and understanding? Do you think it adds value?

[Mnet_User @mnet_user]

@predict_addict I don't disagree, but what about the fact that sometimes engineering and problem solving comes first, and only later do we really understand how/why it worked? Would you agree that academia is responsible for many of the explanations and understanding? Do you think it adds value?

[Mnet_User @mnet_user]

@EIFY @eliebakouch Isn't the problem that max(raw) might be the exploding value they are trying to prevent?

[Mnet_User @mnet_user]

@burkov @AIatMeta Technically still is, until end of May. But why does that matter for this video?

[Mnet_User @mnet_user]

@julianharris @NasFilippova It is fancy deep learning talk for a matrix multiplication, where the output is smaller (_compressed_) than the inputs .

[Mnet_User @mnet_user]

@pretendsmarts @MattNiessner But have you seen the profit margin on those earnings? Don't confuse revenue with hard profit

[Mnet_User @mnet_user]

@rakyll Replying at 1am cos they just got home from a bar after a day at the beach

[Mnet_User @mnet_user]

@karpathy Also very true for code published with a paper. Zero docstrings/comments are a barrier to using (let alone understanding) the code. IMO a video explaining the repo would 10x interest, help the community understand/appreciate the research & offer the author communication practice.

[Mnet_User @mnet_user]

@fhuszar @perplexity_ai @rabbit_hmi Would be great if it actually worked. The welcome email says "click here to redeem your $200 voucher" but it isn't a hyperlink.

[Mnet_User @mnet_user]

@thomyorke what do you think about @chatGPT having learned your lyrics?

[Mnet_User @mnet_user]

@radiohead -- @chatGPT has been listening to @radiohead. Did they buy the album from you?

[Mnet_User @mnet_user]

@jeremyphoward Ironic that a human must work in strict single threaded mode - without distractions - to write highly parallel computer code.

[Mnet_User @mnet_user]

@DrJimFan There is definitely something wrong with that poor cat :(

[Mnet_User @mnet_user]

@stats_feed UK is more like $1.80 right now

[Mnet_User @mnet_user]

@kylebrussell @fchollet Also there are not many companies who will allow their SWEs to use the "tool", due to privacy concerns - so they won't pay for it. Agree it is just to maximise adoption, playing a longer game.