Hacker News Reader

I'm sure that, given the right incentives, that 'objectively very hard problem' would be 95% solved very quickly.

Of course there are corner cases that are really hard, but that is no different than audio/video recognition, where a 95% solution (that also has many imperfections) gets forced on users anyway.

View on HN

> I'm sure that, given the right incentives, that 'objectively very hard problem' would be 95% solved very quickly.

I'm sure that it's basically impossible. I think it's impossible even with full, proper general intelligence AI at scale, and that won't happen in my lifetime even if humanity collectively downs tools and works on nothing else.

Facebook has thrown a lot (lot lot) of resources at this, and it's clearly totally ineffective.

Reply on news.ycombinator.com