| ▲ | crooked-v 3 days ago | |
This article seems to me like a lot of "if you solve all the hard problems, then you'll have a solution". Which is like... yes, and...? | ||
| ▲ | mdp2021 3 days ago | parent | next [-] | |
It's relevant given the apparent conservativeness about LLMs, the non-revolutionary architectural improvements that seemed to place too much of the bets on scaling. | ||
| ▲ | slashdave 3 days ago | parent | prev | next [-] | |
The article doesn't even discuss hard problems. An unfortunate tendency that many in high-tech suffer from is the idea that any problem can be solved with engineering. | ||
| ▲ | jibal 3 days ago | parent | prev [-] | |
It lays out what those problems are and how LLMs don't solve them. | ||