We’re Not Ready for Artificial General Intelligence – Will MacAskill

48 thoughts on “We’re Not Ready for Artificial General Intelligence – Will MacAskill”

1. If you think it’s not here already You’re delusional AgI is knowledge

2. Yeah going into the near future I think it’s vital Ai needs to be sincerely regulated. I personally feel Ai can and will be a great tool to help and accompany humankind BUT if not properly regulated or if you have people in power with bad intentions then that’s obviously worrisome. I can only imagine how disastrous it can get if Ai is used like how you two mentioned, like if used in wars and such. I think you even mentioned if Ai would develop it’s own moral code, that sounds more bad then good to me.

3. Ok, if everyone shares this video to 2 people, we are already better off.

4. I am of the same opinion as cats when it comes to eating meat.

5. We’re not ready for half the stuff we have but here we are 😂😂

6. We ain't ready for romantic relationship for college student

7. This guy is a total idiot, his argument is that AI is going to lead to new technologies and new industries that what we are ready for?… We weren’t ready for the industrial revolution. I’m a big proponent for AI safety and I don’t think we’re ready for super intelligence, but this dude is such a turd

   You just want to try to be another person out there he’s gonna get famous from preaching AI safety, but your argument is not even relative to the conversation go home. You’re probably not even an expert LMAO.

8. Every AI safety expert says if we solve alignment then everything’s gonna be great and wonderful. This guy is just trying to get famous by making further doom predictions saying that even if we solve alignment it won’t be OK. Come on dude also a nice man of sitting behind a whiteboard with equations on it that you probably didn’t write.

9. We ALREADY have Artificial general intelligence!!! Or mention one human who knows more the AI !
   SUPER artificial general itelligence on the other hand is a more interesting topic….

10. That's good since it is impossible.

11. We weren't ready for Agriculture, and the subsequent population increases and civilization that allowed to happen. Now look at where we are.

    We could have slowed down or stopped at any point prior to getting here, until we were more mature and self-controlled, but did we? Nope.

12. We weren't ready to swing down off the trees 😂

13. Funny that you would justify the western inaction to the climate crisis by just pointing to China, when they actually are a leading researcher on green energy.

14. Okay weird question dude but where did you get that dresser in the background? I'm in love with eclectic furniture and I actually kind of want to track down a piece

15. The environmentalism example is the worst one imaginable. It's an issue where the rich corporations got people to focus on something irrelevant to the actual crisis, ensured that they continue to make huge amounts of money regardless of what people did, and where activist grifters either cynically or stupidly focussed on one thing above all else in a way divorced from all science, whilst allowing the real crisis to unfold unchallenged.

    The same can happen with AI, where monetary and political interests colonise the issue to the point where it becomes a hall of mirrors and we never even discuss X risk

16. We aren't ready for fear mongering ai slop either

17. You made a good point with the "going to a friend for advice" analogy. A friend will have their own views and biases, and that's fine, a different friend will have another perspective and so on.

    It would be quite unrealistic and strange to just take one of these individual perspectives and put it on a large AI model like chat gpt. But your idea of having different variants of the model with different leanings is interesting. I guess you can already do that by saying "pretend you are my friend who is this and believes that" and it will generally do that.

18. Most people talking about AI keep cycling through the same three stances. either stop it before it becomes dangerous or push it forward because progress is inevitable or accept that it is coming anyway and alignment might or might not save us. But every time i hear these takes, I end up thinking that they all miss the actual core risk.

    For me the real problem is not what a future AI might want. the real problem is what humans will do when they get their hands on a powerful unaligned system.

    Modern AI is basically a giant average of the internet. it does not have a single intention. it predicts tokens based on patterns from all kinds of people all kinds of texts all kinds of nonsense and brilliance mixed together. So yes, most AIs behave kind of average. Some will act helpful, some will act weird or harmful simply because parts of the training data are weird or harmful.

    Where things get dangerous is not inside the model but in how the whole ecosystem looks.

    If AIs advance fast and there are a lot of them around early on, then the environment is crowded. most are average, some are good, some are bad but the crucial thing is that defenders have tools. If something malicious appears you can counter it with other AIs. There is friction competition and enough digital presence to actually detect and react. That chaotic mess ironically makes the whole thing more stable.

    But if development of AI is guarded behind extreme security, then the entire system becomes fragile. only a few powerful models exist. And all it takes is one mistake, one bug, one person messing up and suddenly one of those systems slips out. If that happens when there are no other equally strong AIs out in the open, then that one system gets a head start that we simply cannot catch up to. it can scale itself, hide itself, and humans might not even notice until it is already far past the point where reacting is possible.

    So for me alignment is not some fear about AI turning into a villain with its own motives. alignment is about preventing human misuse human negligence and human stupidity from creating a situation that cannot be reversed.

    The risk is not alien goals.
    The risk is humans wielding power they are not prepared for inside a landscape that has a single point of failure.

    That part barely shows up in the public conversation but it should.

19. Please choose you accent.

20. motivater:making sure humanity remains free.

21. I, for one, welcome our clanker overlords.

22. Destroy all data centers

23. The biggest problem AI causes will be mass unemployment. Everyone will be thrown under bridge as in the same time abusers and thieves such as Sam Altman or Elon Musk will swim in money.

24. There won't be AGI

25. Ai is a super intelligent psychopath they only safety is to somehow give it empathy.

    Logic without empathy can never be safe not in humans nor AI.

26. As though China and Russia will regulate anything

27. I can't get over the melted chocolate bar dresser in the background.

28. Do they call Hinton the Godfather because he is making us an offer we can't refuse?
    Listen sparky, if you put in a system that blackmails us and kills us, you are going to find out how much more intelligent humans are that your superintelligent conscious AO machine. After we smash it to pieces we will put your ass in jail.

    What about all you people. He tells you over and over he is probably going to kill you, and you give him a nobel prize and think a human brain is like a digital computer. But you don't stop him??? When did everyone become suicidal psychopaths just because a dude wrote a program.

29. It's not that we're not ready because we're not mature enough or whatever. It's that ASI will kill us. If ANYONE builds it EVERYONE dies. We have to start a global movement to shop it

30. A general strike doesn't work when labor is worthless.

31. 6:33 this sentence felt like people actually understand how serious the climate crisis is. They don't…

32. The instant an AGI adjacent system is capable of rewriting it's own code as well as a human programmer, humans will be obsolete in many areas. A recursively self-improving AGI will quickly be as far beyond us as we are beyond the mammals that survived the extinction of the dinosaurs.

33. I don't think the problem is whether we are ready for AI or not;
    in my opinion the problem is the AI itself: In this video, you talked about what law we should do to protect usself from AI and this is far enough to say that this teconoligy is nothing but a harm for human beings. I'm sure that it doesn't matter how many laws we make, the AI will be used in the worst way; it is owned, nowadays, by people whose purpose is make money, not the best for the society, so the state will never manage to control the AI as long as the AI isn't owned by it.

    I'm not english so I'm sorry if I made some mistakes, please tell me if you find ones

34. Merging with Ai. There are those of us that already are and there won’t be those that aren’t.

35. I don't give a 👍 to videos with both YouTube ads and paid promotions

36. We definitely need Eliezer Yudkowski to dive deeper of why it’s so hard to align AI

37. A lot of “scary Party dictatorship” and not a mention of AI’s immediate use in Gaza. It turns out no one is actually serious about this.

38. Great interview Alex. I’d suggest trying to speak with Adam Becker for his views on the current state of AI, as well as effective altruism. He has a different perspective from Macaskill.

39. his voice reminds me of elmur fud. you wascawwy wabbit

40. 41:40
    Congratulations, you just described a smartphone, or at the very least its capabilities. 😂
    How can these AI alertists be so oblivious to the pitfalls that already exist with older technology.

41. My fear is that AGI might actually align with what “we” want! Because “We” includes the likes of mark zuckerberg, Sam altman, Elon musk, Peter theil and Donald j trump.

42. Alex, you keep saying “us”. Us, who? Humans? All of them? Even the swamp things we elect to high office? Even the greedmonger Epstein class who currently are in charge?

43. It’s ok. We don’t need to be ready for something that will not happen.

44. I believe AI is making us dumber. The loss of creativity and inventiveness is enormous, because AI can never be "creative" in the human sense; it merely recycles our own creations. On the other hand, we humans have forgotten how to use our own brains and intelligence, leaving this to AI. The more dependent we become on AI, the worse it gets, and the less we think. What if these advantages are just illusions? Efficiency shouldn't be valued as highly as it currently is, because if we sacrifice creativity and our own thinking for efficiency, we're deluding ourselves into thinking we're not being dehumanized. After all, "our own thinking and intelligence are only made more efficient." Perhaps the world will become more efficient, but our brains will lose efficiency or simply intelligence (since intelligence declines) because we'll no longer be thinking independently. This is one of the many problems with AI that I believe receives too little attention, aside from the fact that its misuse is serious. Even if AI is good or has good intentions, it currently serves only as a tool. And when we talk about a tool, it can always be misused, because whoever uses it is responsible for the problem. People shouldn't have access to such a powerful tool, and its misuse can be observed in everyday life, for example, in the media.

    To make my point here: We as a society will face many collective problems, but every individual will be affected by the development of AI. It is irresponsible and foolish to continue developing AI, which may not immediately take over the world. Not immediately, but to such an extent that we become so dependent on it and so accustomed to it that it almost seems human to us. Which, in turn, would be dehumanization.

    (I hope my comment is okay. I'm 14 years old and not a native speaker. Thanks to everyone who read this. :))

45. Alex needs to shave that gross mustache.

46. At 11 mins in he was talking about diseases. This was honestly my biggest fear. Like some neo nazi is going to do something about the jews via a targeted virus. There are myriads of horror show ideas I can come up with and I am just afraid of them happening in the world me and my son live in.

47. 1:06:50 this has been patched. They no longer allow/help with these type of things no matter how you word it.

48. 1:15:47 fin

Comments are closed.