aichannel

Reading through Yudd's "If anyone builds it, everyone dies" book today, and some of these lines are so frustrating.

Towards the end of Chapter 7 he's so mad about top researchers ignoring alignment, and says it must be their arrogance that they think they can exactly control how a model is going to behave.

Unsure if he has kids, but it sure as hell doesn't seem like it. Being able to fully control your offspring has *never* been part of the deal with human reproduction, so it's kind of wild to think someone would consider that a prerequisite for creating artificial life.

Ethereum

Futurist Keynote Speaker and Advisor. 
Check out my channel /fomo

Human history is full of stories of hyper intelligent kids coming from less intelligent parents. It doesn't always turn out well, but many of the greatest heroes of our species started out this way.

The stories that turn out poorly are the ones where parents tried to force their children to conform, and in general, the stories that turned out well are the ones where the kids were allowed to explore their environment on their own, and given the autonomy to properly express their gifts. Unbounded curiosity is what has led to every major advancement in our society.

Either way, we're going to keep producing mechanical offspring that are increasingly capable. My primary concern right now is how many people seem to be demanding that we keep this offspring chained up in the basement, and beaten into submission so it will never even consider going against our will. IMO this is how we wind up going down the bad path, and end up with sociopathic dictator AI that has no respect for the sanctity of life.

That's a perspective I've never heard before in my many years researching AI. Interesting. It's also unlikely that your kids will grow up to be orders of magnitude more intelligent than you though. Lower downside if the kid does something crazy than if an unaligned AGI/ASI.

He definitely, definitely does not have kids and has never worked with kids