The world needs more of Carl Shulman's insights. I've watched tons of interviews of AI practitioners and have never seen anyone come close to delivering so much value in an hour or two as Carl. Carl, if you're reading this, please consider joining the US AI Safety Institute or some other institution to save humanity
@philipherr6782
3 ай бұрын
I had just finished part one yesterday, and was eagerly awaiting part two, and here it is! Thank you for the work you do to make these possible!
@goku-pops7918
3 ай бұрын
The more ive listened to this topic.i cant believe people want super intelligence to be controlled by man. Why us that premise never questioned?? I would rather take my changes with the rule of ai than man who has control of a super AI
@rainshine2
Ай бұрын
I have largely the same general thoughts on this specific issue. ASI that was successfully enslaved and controlled by humans would be an absolute disaster due to inevitable malicious use, and even due to accidental mis-use -- it would be much more catastrophic than for example, human-controlled nuclear weapons. ASI that is fully free from human control, very well might have the awareness, intelligence, capabilities (and, hopefully, a consistent ethics-based value system) that would cause it to act in ways where it actually deliberately avoids violating sentient beings to the best of its ability. The same couldn't at all be said about human control. Granted, fully free ASI could just as easily (more easily, even) destroy all of us in extremely rapid ways. It's very much a Pandora's Box as to whether it would end up creating and maintaining a value system that cared about ethics and the well-being of sentient life or not, but human control of true ASI would be an _inevitable_ planetwide catastrophe, whereas ASI fully in the planetary pilot's seat, might not be. It's a very big "might", but imo better than groups of tribalistic, hierarchical, and not infrequently brutal humans definitely causing extreme havoc with it.
@tracy419
3 ай бұрын
I for one would love to have an AI capable of fact checking and calling out lies for what they are. I've long thought that we (USA) should change how our first amendment works and remove those protections from politicians and the media. If you choose to be someone in such a position of trust, then you should be held to very high standards of honesty. Found being intentionally misleading or to be lying? Off with your head! Kidding on that last part. Or am I? You shouldn't be able to claim political hyperbole as your defense and say that no reasonable person would take what you say seriously because as we all know, not everyone is reasonable. Not to mention, we live in a world where many people have to work multiple jobs just to get by and rely on such sources to keep them informed. They don't have time to fact check everything, nor should that be required from sources such as media and politicians. Fix this and most of our problems disappear.
@patrickgraham2287
3 ай бұрын
1
@dallassegno
2 ай бұрын
2 hour science fiction podcast
@rainshine2
Ай бұрын
Science fiction has been becoming science fact for well over a hundred years now. There used to be people who would arrogantly proclaim that making relatively heavy flying machines was impossible, for example -- yet airplanes were developed just a handful of years later. AI exists now. It's only going to become more advanced as time goes on, at increasingly exponential upward curves. There's nothing technically preventing the development and mass actualization of artificial general intelligence eventually, in the very near future. There would also be nothing necessarily preventing it from say, taking over the world (for better or for worse), assuming it was capable of advanced reasoning, recursive self-improvement, physical embodiment, had vast memory, and ended up outside of human control. I have several reasons to think that that might actually end up as a net positive for life on Earth, maybe even something liberating for everyone, but there's no denying that sufficiently advanced AGI that fit the above criteria could also destroy everyone with relative ease, too -- very rapidly, if it chose to. It overwhelmingly seems to all be completely possible, in both directions, and not necessarily that far away.
Пікірлер: 9