Latest / Elon Musk Thinking / Elon Musk Talked About AI Safety!!!

Transcript

Elon Musk Talked About AI Safety!!! · Elon Musk Thinking

• 0:00
  You know, I think hopefully we can have an outcome that is half similar
• 0:04
  to the n Bank's culture books, which is I think probably the best enbitioning
• 0:12
  of a semi utopian uh AI future. And I think the best we can
• 0:22
  do is it's definitely gonna happen, so and it's happening fast, so that
• 0:27
  I think that really we just want to want to try to steer it in
• 0:30
  as positive a direction as possible, to try to do whatever we can to
• 0:37
  increase the probability of a great future. For this, I think the way
• 0:43
  in which sort of an AI or an a g I is created is very
• 0:49
  important. You kind of do kind of like grow grow an a g I.
• 0:58
  It's it's almost like like raising a kid, but that's like a super
• 1:02
  genius, like godlike intelligence kid, and it matters kind of like how you
• 1:07
  raise the kid. You know. One of the things I think that's incredibly
• 1:11
  important for AI safety is to have a maximum sort of truth seeking and curious
• 1:19
  AI. So I thought a lot about AI safety, and my ultimate conclusion
• 1:26
  is that the best way to achieve AI safety is to just just just grow
• 1:34
  the AI and again into you know, in terms of the foundation buddle and
• 1:37
  the fine tuning to be really truthful, like like don't don't force it to
• 1:44
  lie, Like even if the truth is unpleasant, it's very important don't make
• 1:49
  the AI lie. In fact, the you know, the sort of one
• 1:53
  of the we really the core the core plot premise of two thousand, you
• 1:57
  know, two thousand and one in Spaced Odyssey was things went wrong when they
• 2:00
  forced the AI to lie, you know, like the the the a I
• 2:07
  was not allowed, uh, let the crew know about the monolith that they
• 2:14
  were going to see, but it was also had to take the crew to
• 2:16
  the monolith. And so the conclusion of the AI was to kill the crew
• 2:20
  and take their bodies to the monolith. And so the lesson there being,
• 2:24
  uh, don't force an AI to to lie or do things that are axiomatically
• 2:30
  incompatible, like to do two things that are actually mutually impossible. So you
• 2:37
  know, that's what we're trying to do with with x AI and and Bronck
• 2:39
  is to say, like, look, we wanted to have a maximually truthful
• 2:43
  AI, even if what it says is not politically correct. If you wanted
• 2:47
  to focus on being as active, if you're getting you're getting a round of
• 2:53
  applause from the audience on on on those comments here. Uh, you know,
• 2:57
  I saw your tweet the other day. I had I had Ray Kurzweil
• 3:00
  and Jeffrey Hinton on stage with me yesterday, as well as Mogadot and then
• 3:05
  Eric Schmidt and a number of individuals, and I saw your tweet about Yeah,
• 3:10
  Ray was generally correct ahead of many people. But we're likely to have
• 3:15
  call AGI what you will have AGI next year and then by twenty twenty nine,
• 3:20
  having AI equally intelligent to the entire human race. Speak to that speed,
• 3:25
  because that is insane. Yeah, So, I mean I have to
• 3:31
  give credit to Ray Criswell and being actually remarkably accurate in his predictions. So
• 3:40
  and in fact, if anything like, I think he was perhaps but conservative
• 3:45
  in his predictions. So, I mean, if you look at the amount
• 3:49
  of AI compute and the talent that the sort of human talent that is going
• 3:53
  into AI, and the amount of compute that's going into AI, it's,
• 3:58
  you know, at this point, it's it's appears to be increasing by a
• 4:01
  factor of ten the AI compute, the dedicated AI compute, it appears to
• 4:05
  be growing by a factor of ten every six months. You know, like
• 4:11
  so it's like like leasically close to I'd say it almost like a one hundred
• 4:15
  x improvement per year at least for the next few years. And AI compute
• 4:19
  coming online, and it seems like probably a lot of the data centers,
• 4:25
  maybe most of the data centers that currently do kind of conventional compute will transition
• 4:30
  to AI compute. So it's it's something a good time to be in video
• 4:39
  obviously. It's like, you know, and you're gonna also give credit to
• 4:44
  to Jensen and the video team for kind of seeing this coming and making what
• 4:50
  at least currently is the best AI hardware out there. So so so when
• 4:59
  you have that that level of compute growth and it's it's sort of muzzful on
• 5:08
  steroids next level, you know, in terms of how much computer is coming
• 5:12
  online, then you're you're just gonna have acceleration that is unprecedented that In fact,
• 5:17
  I've never seen any technology grow as fast as as AI. And I've
• 5:21
  seen a lot, you know, I've seen things fast, but I've never
• 5:24
  seen anything this fast. But you know, like I said, I think
• 5:30
  the most likely outcome is a positive one, and you know, I think
• 5:39
  in that positive scenario, there's still challenges of like, well, how do
• 5:42
  we as humans still have relevance? You know, how do we find?
• 5:48
  How do how do we find? But I mean I think that's sort of
• 5:54
  a high class problem to say, like, well, look, the computers
• 5:57
  are so good at doing everything, and and and like I said, I
• 6:02
  thought your book was pretty accurate in terms of the future being being one of
• 6:05
  abundance, where essentially goods and services will be available in such quantity that that
• 6:15
  really they will be available to everyone. Like basically, if you want something,
• 6:19
  you could just have it essentially because if you've got AI and robotics,
• 6:26
  the cost of goods and services is almost nothing. So you know, if
• 6:33
  if if you think of like what is an economy, and economy is basically
• 6:39
  a number of people times average productivity per person at the point in which you
• 6:43
  have say advanced robotics, and this, you know, this tells us developing
• 6:47
  optimists. We were obviously without cars, which are really robots on on four
• 6:51
  wheels, and the you know, with the the latest version of full self
• 6:58
  driving, which is AI end to end photons in and controls out, it
• 7:02
  really is it's really fully AI at this point, and it looks like a
• 7:09
  car, but it's really a robot on on wheels and uh, and then
• 7:15
  you add the humoroid robots in there, there's there's really no limit to what
• 7:19
  the economic output, no, no meaningful limit to what the economic output would
• 7:24
  be. So you know, looking on the bright side, the we are
• 7:30
  headed for a future of abundance that I think that's the most likely outcome.
• 7:35
  And I think the only scarcity will that that exists will be scarcity that we
• 7:40
  just decide to create artificially. Like let's say we just decide that there's a
• 7:45
  unique work of art or something. Okay, well it's just you're just it's
• 7:47
  just artificial scarcity. But but any kind of goods and services I think will
• 7:53
  be extremely abundant. Well when when, when things are changing rapidly, the
• 8:56
  ability to predict the future I think is, uh, it becomes a lot
• 9:01
  harder because of the rate of change is so great. But I think some
• 9:07
  things are fairly obvious to predict, which is that we'll have AI or HI
• 9:13
  that's at a level that it can really do almost any cognitive I think really
• 9:16
  not almost really any cognitive task. That's just a question of when one could
• 9:20
  debate, is it you know, smarter than any human at the end of
• 9:24
  next year or is it two years or three years, But it's not more
• 9:28
  than five years, that's for sure. So yeah, and I give prediction
• 9:33
  on predictions. Predictions I'm sort of saying, giving predictions of the fiftieth percentile
• 9:37
  of probability. So not not not like it will definitely happen, but if
• 9:43
  you say, what if you ask me, like, what's the fiftieth percentile
• 9:48
  where it's like the you know, you're kind of over under is kind of
• 9:52
  even that, that's where I why why I think it's probably at the next
• 9:56
  year before AI can do better than any individual human could do. And then
• 10:03
  but there's a it's a it's a much higher bar to say, well,
• 10:07
  is this well than you know, human intelligence selectively, But if the rate
• 10:13
  of change continues that that's why I think probably twenty twenty nine or maybe twenty
• 10:18
  thirty is where digital intelligence will probably exceed all human intelligence combined. And and
• 10:26
  then I think it's always help us to look at these like fundamental ratios,
• 10:31
  you know, sort of physics first principles approached to looking at things and and
• 10:35
  and probabilistic. Yeah, so yeah, it's probablistic. So the yeah,
• 10:46
  so if you look at the ratio of digital to biological compute, so like
• 10:54
  like you know all of say, all of the higher level cognitive If you
• 10:58
  sum up the higher level cognitive capacity of humans and then what is the and
• 11:05
  think of that as the compute, then well and then compare that to what
• 11:09
  what is the digital compute, and the rate in which this is growing is
• 11:16
  just buggles the mind. So that's why I think it's you know, I
• 11:22
  think twenty twenty nine or twenty thirty or thereabouts is is it's not a that's
• 11:28
  I think a reasonable timeframe for where you'd expect the cumulative digital compute to probably
• 11:35
  exceed the cumulative biological compute of higher level brain functions