kingstnap

28 ms·

481.
by kingstnap 1y ago
The recent Apple "LLMs can't reason yet" paper was exactly this. They just tested if models could run an exponential number of steps. Of course, they gave it a terrible clickbait title and framed the question and graphs incor
482.
by kingstnap 1y ago
EVs are better suited to most commute patterns. The ICE would be the compliment, and the EV, the daily driver. With better charging infrastructure, that most turns into virtually all.
483.
by kingstnap 1y ago
These articles and papers are in a fundamental sense just people publishing their role play with chatbots as research. There is no credibility to any of it.
484.
by kingstnap 1y ago
Because they have unrealistic targets so they make up fake uptime numbers. 99.999% would mean not even having an hour of downtime in 10 years. I remember reddit being down for like a whole day or so and they claimed 99.5% in that month.
485.
by kingstnap 1y ago
Being a first-party source and reliability are orthogonal concepts. But yes, he doesn't need to bring receipts; he is the receipt. He can take the number of queries they ran and divide it by the amount of electricity they bought. Or ma
486.
by kingstnap 1y ago
> Altman did not offer any evidence for these claims and failed to mention where his data comes from. Presumably, as CEO, he is a first-party source. Also, I mean, water usage is relative. How much water does a person who chats all day u
487.
by kingstnap 1y ago
Writing code that is mindful of power draw is the same as writing fast code. Race to yield to the os as soon as you can. Minimize data movement. Use efficient algorithms. Be mindful of abstraction overhead, etc.
488.
by kingstnap 1y ago
Proof of Work provides scarcity that makes crypto work. Stake just removes the physical scarcity so it fundamentally can't work. I think proof of storage or something would be more interesting as a (partial) solution to the crypto folk