My dad uses Google Maps, and he mentioned that it seems to be getting worse. Like, giving him directions that are obviously worse than alternatives. Has anyone else here experienced this?
When AI gets applied to robot bodies, real world results will be able to trim out bad knowledge. Currently because AI only feeds on internet content, all the AI has to eat is human content and AI content.
AI will drift away from accuracy until it gets embodied at which point it will start to get more accurate.
Real world experience can help, but what we have now is also too stupid to recognize when it’s succeeding or failing. It just greedily gobbles up inputs and feedback indiscriminately.
There’s currently no way to know if the necessary advancement, to advance independently of humans, is 2 years or 2000 years away.
Even so, nature tells us that advancement probably isn’t coming at all. It’s not needed, so long as there are billions of humans available to partner with.
