00:00Dr. Jampolsky, what do you think? Can AI be ethical? And does it currently, as Amjad has suggested, have a Western bias?
00:09So usually when people realise they cannot make progress on a hard problem of AI safety, meaning controlling advanced agencies, they switch to an easier problem of ethics.
00:18Ethics is something humans have been arguing about for millennia. Philosophers have never solved it. It's region-specific.
00:25What is ethical here is unethical there. But we can certainly get AI to mimic human level of ethics.
00:31They can repeat the same biases, and we can remove that bias. So that's quite possible, doable.
00:37It seems that in the marketplace, there are two approaches. One is this very strict filtering.
00:42The examples given is the failure of that approach. The alternative is what Elon Musk is doing with Grok.
00:48Kind of do whatever you want, have fun with it. It's not a big deal.
00:51And so that seems like both of those models are very commercially successful.
00:56There is no problem with doing it either way.
01:00And there is going definitely to be a competition between those who provide very, let's say, kid-friendly models
01:07and those who provide models who tend not to say no to your requests unless you're asking for biology, synthetic weapons, and things like that.
Be the first to comment