Every story and comment from the Hacker News API, in a layout borrowed from 2001.
A comment on Have we screwed up artificial intelligence?.
- As Papers without Code website presented: many algorithms in papers are manually super-optimized to perform state-of-the-art on benchmarks
- Much of ML work outside of top academia and FAANG lacks rigour or doesn't compare results with "dumb" baselines.
- Many companies (e.g. Google, FB) have not devised a proper recourse procedure for mistakes their "AI" policing bots make (and WILL always make).
However:
- Weren't it for commercialization of ML, we would not have ML accelerators of some sort in almost all modern hardware.
There are no comments on this comment yet.
- As Papers without Code website presented: many algorithms in papers are manually super-optimized to perform state-of-the-art on benchmarks
- Much of ML work outside of top academia and FAANG lacks rigour or doesn't compare results with "dumb" baselines.
- Many companies (e.g. Google, FB) have not devised a proper recourse procedure for mistakes their "AI" policing bots make (and WILL always make).
However:
- Weren't it for commercialization of ML, we would not have ML accelerators of some sort in almost all modern hardware.