Latest / Elon Musk Thinking / Latest: Elon Musk, Jensen Huang, Tom Brown Discuss the AI Revolution!!!

Transcript

Latest: Elon Musk, Jensen Huang, Tom Brown Discuss the AI Revolution!!! · Elon Musk Thinking

• 0:00
  Speaker 1: Latest interview of elon musk.
• 0:03
  Speaker 2: Hey team oh gee are we late.
• 0:11
  Speaker 1: Guys good to be here go america go america team
• 0:15
  Speaker 1: america yes indeed so america right brothers first man on
• 0:22
  Speaker 1: the moon the internet.
• 0:23
  Speaker 3: The two of us right brothers i like.
• 0:25
  Speaker 2: That indeed we've essentially led every major technological revolution, which
• 0:32
  Speaker 2: has been to the enormous benefit of America, ordinary Americans,
• 0:36
  Speaker 2: where I guess we're now calling this superintelligence, which I
• 0:39
  Speaker 2: think is actually good branding. And so let's talk about
• 0:43
  Speaker 2: how America wins this, and we make that winning good
• 0:46
  Speaker 2: for ordinary Americans.
• 0:48
  Speaker 3: Well, excuse me, Diet Coke.
• 0:53
  Speaker 4: Sorry.
• 0:55
  Speaker 5: So first of all, Artificial superintelligence is a complete reinvention
• 1:02
  Speaker 5: of the computing stack. Everything from the way the processors,
• 1:06
  Speaker 5: the chips are designed and the way they operate to
• 1:09
  Speaker 5: the way software is now trained and developed to the
• 1:14
  Speaker 5: way that the applications are able to perform what they're
• 1:17
  Speaker 5: able to do and how they perform.
• 1:19
  Speaker 3: Every single layer of the computing stack has been reinvented.
• 1:23
  Speaker 5: This way of doing computing Unlike previous ways of retrieving information,
• 1:28
  Speaker 5: this is a generative approach to computing.
• 1:30
  Speaker 3: And therefore, in order for us to succeed, the first
• 1:33
  Speaker 3: thing we need is energy.
• 1:35
  Speaker 5: Because we need the energies to run the compute, to compute,
• 1:39
  Speaker 5: run the algorithms, the algorithms then feed the applications. And
• 1:44
  Speaker 5: so at the foundational layer is energy. And from that energy,
• 1:48
  Speaker 5: we build the land power and shell and that hosts
• 1:51
  Speaker 5: the infrastructure, so on and so forth. And so the
• 1:53
  Speaker 5: way we win, of course, we are world-class in chip design.
• 1:57
  Speaker 5: We're world-class in algorithm development. And so one of the
• 2:03
  Speaker 5: things that we need to do is just be vigilant
• 2:06
  Speaker 5: about enabling power delivery and power generation to happen. And
• 2:10
  Speaker 5: President Trump has been visionary in being pro-energy. Without the
• 2:15
  Speaker 5: pro-energy vision of President Trump, it would have been impossible
• 2:17
  Speaker 5: for us to build the data centers on top of it.
• 2:20
  Speaker 3: Now, of course... These data centers are coming online so quickly.
• 2:24
  Speaker 5: We're coming into a lot of neighborhoods and a lot
• 2:28
  Speaker 5: of communities. We have to do a better job in
• 2:31
  Speaker 5: partnering with the communities and bring them along and make
• 2:35
  Speaker 5: them part. of the prosperity and the success of this
• 2:38
  Speaker 5: industrial revolution.
• 2:39
  Speaker 3: But that's how we win. We've got to build the
• 2:42
  Speaker 3: world's best chips.
• 2:43
  Speaker 5: We have to have lots and lots of energy, create
• 2:46
  Speaker 5: incredible algorithms. And then lastly, make sure that America is
• 2:50
  Speaker 5: enthusiastic about the technology so that we diffuse it into
• 2:53
  Speaker 5: all of these various industries, from manufacturing to healthcare, space exploration.
• 2:59
  Speaker 3: So on and so forth.
• 3:00
  Speaker 4: Awesome.
• 3:01
  Speaker 3: Elon, how do we win? How do we keep winning?
• 3:06
  Speaker 6: Well, we need a lot of power, a lot of
• 3:12
  Speaker 6: energy generation. So scaling energy, I think, is one of
• 3:19
  Speaker 6: the biggest things. You've got to scale energy, you've got
• 3:21
  Speaker 6: to scale chip production. And I think we're winning on software,
• 3:27
  Speaker 6: on anything intellectual or digital. But in the long run,
• 3:32
  Speaker 6: we need to be able to generate enough power to
• 3:37
  Speaker 6: compete with, I think, essentially China is the biggest competitor.
• 3:41
  Speaker 6: And China has about three times the electricity production of
• 3:45
  Speaker 6: the United States. And I think we'll figure out how
• 3:48
  Speaker 6: to make ships. So really the central challenge, I think,
• 3:55
  Speaker 6: is long-term is going to be how do you generate
• 3:57
  Speaker 6: enough power and how do you ensure that we have
• 4:01
  Speaker 6: enough logic and memory fabrication capacity in the United States
• 4:04
  Speaker 6: or in a safe region Well, you are.
• 4:07
  Speaker 3: We need the Wright brothers on the job.
• 4:09
  Speaker 4: Yes.
• 4:10
  Speaker 2: Well, it turns out the Wright brothers are on the job. Elon,
• 4:12
  Speaker 2: if you bring on 10 gigawatts, which SpaceX has discussed publicly,
• 4:18
  Speaker 2: that'll be about 25% of all the power added in America.
• 4:21
  Speaker 2: And I know you have literal rocket engineers working on
• 4:25
  Speaker 2: bringing on.
• 4:25
  Speaker 3: The power, helping make the turbines. So can you talk
• 4:29
  Speaker 3: about that?
• 4:31
  Speaker 4: Yeah, it's... I mean, just to frame this.
• 4:34
  Speaker 6: The average power consumption in the United States is about
• 4:37
  Speaker 6: 500 gigawatts. So every 5 gigawatts of incremental steady state
• 4:41
  Speaker 6: production is a 1% increase in the power used. And
• 4:47
  Speaker 6: I think actually will materialize to be approximately 1% increase
• 4:51
  Speaker 6: in GDP. Because the intelligence per watt keeps increasing. So
• 4:56
  Speaker 6: if you just plot on a curve, how is the
• 5:00
  Speaker 6: useful intelligence per watt increasing How is the, both at
• 5:04
  Speaker 6: the hardware level, at Jensen's, GPUs keep getting better in
• 5:08
  Speaker 6: terms of the compute per watt, and the algorithms also
• 5:14
  Speaker 6: get better in getting more intelligence per watt. So that's
• 5:21
  Speaker 6: maybe an interesting note.
• 5:23
  Speaker 4: I would bet.
• 5:24
  Speaker 6: Anyone that 1% increase in power usage corresponds to roughly
• 5:27
  Speaker 6: 1% increase in GDP, and so 10 gigawatts would be
• 5:30
  Speaker 6: a 2% increase in GDP.
• 5:31
  Speaker 4: Thank you. with the Rubens, I think.
• 5:36
  Speaker 3: Well, let's do the simple math.
• 5:38
  Speaker 5: So each gigawatt can probably generate about, call it, $ 40,
• 5:43
  Speaker 5: $ 50 billion of economic output each year.
• 5:48
  Speaker 4: Yeah, I mean, 1% would be, yeah, I mean, yeah.
• 5:52
  Speaker 3: 1% of the United States would be.
• 5:54
  Speaker 4: 300 billion.
• 5:55
  Speaker 3: Yeah, 300 billion. And so 50, 1 sixth. 1% of the... 1%.
• 6:02
  Speaker 6: Increase, so 5 gigawatts would be, yeah, well, 5 gigawatts
• 6:05
  Speaker 6: would be $ 300 billion.
• 6:06
  Speaker 3: Yeah, right.
• 6:07
  Speaker 4: Exactly, $ 60 billion a gigawatt. Yeah, that's right.
• 6:10
  Speaker 3: The math, math. Yeah, yeah. That's what we're good at.
• 6:14
  Speaker 4: Yes, yes. Yeah. So that's quite a.
• 6:19
  Speaker 5: Not good at too many other things, but.
• 6:23
  Speaker 4: Yeah, I mean, it's going to increase quite dramatically year
• 6:26
  Speaker 4: over year. So, yeah, I mean.
• 6:33
  Speaker 6: From a SpaceX standpoint, we're looking at long-term, how do
• 6:36
  Speaker 6: we access a lot more energy than could be accessed
• 6:40
  Speaker 6: on the ground, and that's with space. So orbital compute
• 6:44
  Speaker 6: is going to be a very big deal. In fact,
• 6:47
  Speaker 6: and Google agrees. They just actually sent up, or are
• 6:51
  Speaker 6: either about to or soon about to send up some TPUs.
• 6:54
  Speaker 6: I think there was an H100 that was put into
• 6:58
  Speaker 6: orbit like last year, I think.
• 7:00
  Speaker 4: Mm-hmm. So it does work in space.
• 7:04
  Speaker 6: And SpaceX is aiming, together with Tesla, to do 200
• 7:10
  Speaker 6: gigawatts of solar production per year. And if brought to space,
• 7:16
  Speaker 6: you get the nameplate or better on the solar power,
• 7:19
  Speaker 6: whereas if it's on the ground, you're going to get
• 7:21
  Speaker 6: somewhere between a fifth and an eighth of the solar.
• 7:24
  Speaker 4: Power generation, and you need enormous batteries.
• 7:26
  Speaker 6: So whereas it's essentially always sunny in space, so.
• 7:31
  Speaker 5: And right now is a great time to do it
• 7:33
  Speaker 5: because the economic returns of these.
• 7:36
  Speaker 3: Computing systems are so good that the.
• 7:40
  Speaker 5: Launch cost, although it's still quite high, in terms of
• 7:44
  Speaker 5: the overall economic benefit, is quite affordable.
• 7:47
  Speaker 4: Absolutely.
• 7:48
  Speaker 5: This is the perfect time to do it. Putting NVIDIA
• 7:50
  Speaker 5: Compute in space is the best way to get there.
• 7:52
  Speaker 2: Yeah, climbing the Kardashev scale with American power and American
• 7:55
  Speaker 2: chips on American rockets, I love it.
• 7:57
  Speaker 3: Yeah.
• 7:58
  Speaker 2: So, Jensen, you said something that I thought was important
• 8:01
  Speaker 2: about how when we bring the energy to all these
• 8:05
  Speaker 2: small towns, we're re-industrializing America, we're revitalizing small towns, we're
• 8:11
  Speaker 2: modernizing the grid. These are things that essentially every American
• 8:16
  Speaker 2: politician has wanted for.
• 8:17
  Speaker 4: The last 35 years.
• 8:19
  Speaker 2: What would you say, what are some specifics you can
• 8:23
  Speaker 2: give about how data centers... are good for ordinary Americans
• 8:28
  Speaker 2: in the communities.
• 8:29
  Speaker 3: And how they're getting better for those communities. Yeah, we
• 8:31
  Speaker 3: used to call them data centers.
• 8:33
  Speaker 5: We call them data centers because that's where we hold data.
• 8:36
  Speaker 5: But the things that we're building now, these are really
• 8:39
  Speaker 5: SI factories, super intelligence factories.
• 8:41
  Speaker 3: And the reason why it's a factory is because.
• 8:43
  Speaker 5: You're producing something of economic value.
• 8:45
  Speaker 3: You're just not storing it.
• 8:46
  Speaker 5: And so these SI factories are real factories. It's like
• 8:52
  Speaker 5: a big, giant power generator in real factories. It's got
• 8:55
  Speaker 5: a lot of, computing and networking in there, power generation,
• 8:58
  Speaker 5: cooling systems.
• 8:58
  Speaker 3: It's complicated things.
• 9:00
  Speaker 5: And so we're building out, you know, just as the
• 9:03
  Speaker 5: Internet today allows you to find anything anywhere, in the
• 9:07
  Speaker 5: future with these AI, excuse me, SIA infrastructure, SIA factories,
• 9:12
  Speaker 5: you'll be.
• 9:12
  Speaker 3: Able to know anything and do anything from anywhere.
• 9:15
  Speaker 5: And so now we're building this infrastructure out and, you know,
• 9:20
  Speaker 5: call it across all of the United States, somewhere between 10,000,
• 9:24
  Speaker 5: 20 gigawatts a year, let's just estimate, that creates just,
• 9:29
  Speaker 5: you know, probably about a million jobs in building out
• 9:32
  Speaker 5: these data centers. Don't forget, you've got to build the
• 9:35
  Speaker 5: power generation plants, you've got to build the data centers,
• 9:37
  Speaker 5: you've got construction, you've got, you know, power cooling, you've got,
• 9:40
  Speaker 5: you know, pipe fitters. And so, this is the first
• 9:43
  Speaker 5: time in probably, as you mentioned, 35, I was going
• 9:46
  Speaker 5: to say 50 years, first time in 50 years, we
• 9:49
  Speaker 5: are re-industrializing the United States, creating a whole bunch of
• 9:52
  Speaker 5: blue-collar jobs and, I think we want an economy that
• 9:59
  Speaker 5: has a shape that includes white-collar jobs and blue-collar jobs,
• 10:03
  Speaker 5: not just white-collar jobs. And so I think that this
• 10:07
  Speaker 5: is our great chance of reshaping our economy and bringing
• 10:11
  Speaker 5: that type of builders and craft back into the United States.
• 10:15
  Speaker 4: Absolutely.
• 10:16
  Speaker 3: And I mean, it feels like what.
• 10:20
  Speaker 2: Colossus and Macro Harder have done have been amazing, the
• 10:24
  Speaker 2: communities around them, the jobs they've created.
• 10:26
  Speaker 3: How do you think about that?
• 10:28
  Speaker 6: Yeah, the economic impact, the positive economic impact has been tremendous.
• 10:34
  Speaker 6: I mean, there's actually a job shortage in the region,
• 10:38
  Speaker 6: whereas there used to be quite a bit of unemployment,
• 10:41
  Speaker 6: and now there's over-employment, if you will. There's more jobs
• 10:47
  Speaker 6: than there are people. and the tax benefits for the
• 10:52
  Speaker 6: community have been tremendous. I think we might have doubled
• 10:55
  Speaker 6: their tax budget or something quite tremendous. We do small
• 11:00
  Speaker 6: things that are helpful, like provide half-price Starlink access. We're
• 11:07
  Speaker 6: building a water recycling plant that's a $ 250 million water
• 11:10
  Speaker 6: recycling plant. I think you do need to do things
• 11:13
  Speaker 6: for the community, and if you do so, they will reciprocate.
• 11:17
  Speaker 4: Awesome.
• 11:17
  Speaker 5: Now, one of the things that people forget is that
• 11:20
  Speaker 5: for the first time in probably forever, that we have
• 11:24
  Speaker 5: to market forces to support the investment of sustainable energy.
• 11:29
  Speaker 5: You give me an example of a fusion, fission, small
• 11:33
  Speaker 5: nuclear reactor, large reactor. Almost every form of sustainable energy
• 11:38
  Speaker 5: is being funded today. And so we have the opportunity
• 11:41
  Speaker 5: to fund the future of energy without the government subsidizing, without, right?
• 11:47
  Speaker 5: I mean, you give me an example of a great
• 11:50
  Speaker 5: energy battery storage idea.
• 11:52
  Speaker 3: It's going to get funded.
• 11:53
  Speaker 4: Awesome.
• 11:54
  Speaker 2: By the way, on behalf of all Americans, I want
• 11:56
  Speaker 2: to thank both of you. I think combined you have
• 11:58
  Speaker 2: created more than $ 5 trillion of wealth for ordinary Americans
• 12:02
  Speaker 2: in their 401ks, tens of thousands, maybe hundreds of thousands
• 12:05
  Speaker 2: of jobs directly. many tens of billions of wages, so
• 12:09
  Speaker 2: thank you.
• 12:10
  Speaker 3: I think Jens has done.
• 12:17
  Speaker 4: Five trillion by himself, actually.
• 12:19
  Speaker 3: I'm sorry?
• 12:20
  Speaker 4: I think Jens has done five trillion by himself.
• 12:23
  Speaker 6: I think I'm around three and a half or something.
• 12:28
  Speaker 2: I was trying to be conservative. So you guys both
• 12:32
  Speaker 2: had a great Monday. I was watching the Starship launch.
• 12:37
  Speaker 2: first orbital Starship launch while I was reading about the
• 12:41
  Speaker 2: open agent safety framework from NVIDIA, which was an exciting morning.
• 12:46
  Speaker 2: So maybe you could each talk about each of those
• 12:49
  Speaker 2: because we need Starship to get those 200 gigawatts of
• 12:53
  Speaker 2: American solder with American chips, American superintelligence into orbit.
• 12:59
  Speaker 3: I'll talk about his and you talk about mine. No,
• 13:02
  Speaker 3: I'm just kidding.
• 13:04
  Speaker 5: So let's see. As you know, we want to advance AI,
• 13:09
  Speaker 5: advance SI. I'm going to get good at this. We're
• 13:11
  Speaker 5: going to advance SI in a responsible and optimistic way.
• 13:17
  Speaker 5: And we know that the future, we strongly believe that
• 13:20
  Speaker 5: the future that SI is going to enable is a
• 13:23
  Speaker 5: future that we want to be part of.
• 13:24
  Speaker 3: And so we're optimistic about that. And that's the reason
• 13:26
  Speaker 3: why we're driving so hard.
• 13:27
  Speaker 5: On the other hand, we have to make sure that
• 13:28
  Speaker 5: we develop this technology safely and securely.
• 13:32
  Speaker 3: There are several different parts of it.
• 13:33
  Speaker 5: While you're training the model, that's generally a fairly safe environment.
• 13:39
  Speaker 5: But while you're evaluating the model, while you're testing the model,
• 13:42
  Speaker 5: you have to make sure that the agentic system, the
• 13:45
  Speaker 5: model itself, is highly contained. That containment system has to
• 13:50
  Speaker 5: be really secure. And the second thing is that while
• 13:54
  Speaker 5: you're testing it, you want to monitor that even though
• 13:58
  Speaker 5: you have it contained, You should never trust that it's contained,
• 14:01
  Speaker 5: and so you want to monitor it in real time.
• 14:04
  Speaker 5: And you want to ideally monitor it with another chip
• 14:08
  Speaker 5: so that it is outside, if you will, called out
• 14:11
  Speaker 5: of band, outside the environment, outside the sandbox of the
• 14:15
  Speaker 5: agent itself. And so you want to contain and you
• 14:18
  Speaker 5: want to monitor. Once you're done evaluating in this way,
• 14:22
  Speaker 5: when you deploy, you want to use a very similar system.
• 14:25
  Speaker 3: And this is no different.
• 14:26
  Speaker 5: Than just about any digital, any human employee that you hire.
• 14:31
  Speaker 5: You give them a containment of rights and tools and access,
• 14:34
  Speaker 5: and then you monitor on a regular basis, on a
• 14:36
  Speaker 5: continuous basis. And so in the case of agents, we
• 14:39
  Speaker 5: just have to make sure containment and monitoring has to
• 14:42
  Speaker 5: be done in the most secure of ways. And we
• 14:45
  Speaker 5: introduced two technologies. One's called Open Shell. It's a containment system.
• 14:51
  Speaker 5: It's basically kind of like a browser system For websites,
• 14:55
  Speaker 5: it's now essentially a browser of agents. And that browser
• 14:59
  Speaker 5: keeps the agent isolated from the rest of the computer,
• 15:03
  Speaker 5: and the rights that you give it are only the
• 15:05
  Speaker 5: rights it should have. And then the monitoring system is
• 15:09
  Speaker 5: a new chip called Bluefield, and it keeps an eye
• 15:12
  Speaker 5: on everything that the agent is doing at all times.
• 15:15
  Speaker 5: And if it does anything that's out of bounds and
• 15:18
  Speaker 5: against the policy that has been given, then it escalates
• 15:23
  Speaker 5: as far as you need it to escalate.
• 15:24
  Speaker 2: And I thought it was a great analogy in the
• 15:27
  Speaker 2: technical blog where you said in the early days of Internet,
• 15:29
  Speaker 2: everybody was worried about your credit card could get stolen.
• 15:32
  Speaker 2: We made the Internet safe, and that's why we have Amazon,
• 15:35
  Speaker 2: we have Netflix, we have Google, we have X. We
• 15:39
  Speaker 2: have so many great things because we made the Internet safe.
• 15:42
  Speaker 2: Now we're doing the same with AI.
• 15:45
  Speaker 3: And safety is not.
• 15:47
  Speaker 5: Inconsistent with advancing the technology. Safety and advancing the technology
• 15:51
  Speaker 5: is one and the same thing. It's no different than
• 15:55
  Speaker 5: when you have autopilot. Not only is it advanced technology,
• 15:58
  Speaker 5: it's also safer technology. And so you advance the technology
• 16:01
  Speaker 5: in that way. The two are not in conflict. The
• 16:05
  Speaker 5: idea that if you advance capability, it somehow... compromises safety.
• 16:10
  Speaker 5: If you have manned safety, somehow you fall behind.
• 16:13
  Speaker 3: That's just, you know, those are false choices.
• 16:15
  Speaker 4: Yeah, safety lets us accelerate.
• 16:17
  Speaker 2: But Elon, if the Starship's going to be a big
• 16:19
  Speaker 2: part of those 200 gigawatts, what can you tell us
• 16:22
  Speaker 2: about Monday's flight?
• 16:23
  Speaker 3: What you learned? How are you feeling about the program?
• 16:26
  Speaker 4: I think the Starship is going great.
• 16:28
  Speaker 6: We took Starship to orbit for the first time and
• 16:32
  Speaker 6: we deployed Starlink B3 satellites, which are enormous satellites. I mean,
• 16:36
  Speaker 6: these are Satellites with a wingspan of 737.
• 16:40
  Speaker 4: So really enormous.
• 16:44
  Speaker 6: In fact, it's the largest payload delivered to orbit since Skylab.
• 16:48
  Speaker 6: So now we expect Starship to ultimately, next year I think,
• 16:54
  Speaker 6: probably reach a weekly or twice-weekly cadence.
• 16:58
  Speaker 2: We've got not Tom Brown here.
• 17:01
  Speaker 3: I know him by his X-handle. Welcome.
• 17:03
  Speaker 4: Thank you.
• 17:05
  Speaker 6: Good to see you guys again.
• 17:11
  Speaker 7: My car took a wrong turn, so you guys got
• 17:13
  Speaker 7: here faster than me.
• 17:14
  Speaker 3: Impressive. I'm glad you made it. Elon, please continue.
• 17:20
  Speaker 2: We're talking about how Starship is going to let us
• 17:22
  Speaker 2: get to 200 gigs of American solar with American chips
• 17:25
  Speaker 2: with American superintelligence.
• 17:28
  Speaker 6: Starship is going to do a lot of things. It's
• 17:30
  Speaker 6: intended to carry enough payload to the moon and Mars
• 17:34
  Speaker 6: to build a self-growing city on the moon and Mars.
• 17:37
  Speaker 6: So I think as a rough order of magnitude, you
• 17:40
  Speaker 6: need about a million tons to deliver to the surface
• 17:43
  Speaker 6: of the moon to create a self-growing city or a
• 17:47
  Speaker 6: similar number for Mars. And then in addition, it will
• 17:51
  Speaker 6: launch at least 300 gigawatts a year of AI compute.
• 17:55
  Speaker 6: 200 gigawatts a year would be a 40% increase annually
• 18:00
  Speaker 6: in the United States total energy consumption. So that's quite
• 18:04
  Speaker 6: a big number. And it might end up being a
• 18:07
  Speaker 6: terawatt one day. So this is serious tonnage.
• 18:12
  Speaker 3: Awesome.
• 18:13
  Speaker 2: And Tom, Anthropic, your company, your co-founder of Anthropic, you
• 18:18
  Speaker 2: announced Opus 5.5 in the last several days. And I
• 18:22
  Speaker 2: thought maybe we'd go to the exponential that we are on.
• 18:26
  Speaker 2: And I think it's helpful to ground What can Opus 5.5,
• 18:29
  Speaker 2: and maybe it's for everybody, what can Opus 5.5 do
• 18:32
  Speaker 2: to make the lives of ordinary Americans better today?
• 18:36
  Speaker 3: And where are we going to be in a year?
• 18:39
  Speaker 7: Yeah, so Opus 5.5 is a model that I think
• 18:44
  Speaker 7: just continues the normal progress that we've had of getting
• 18:48
  Speaker 7: kind of more IQ points for each of these super
• 18:51
  Speaker 7: intelligences with each model release. And so the main things
• 18:57
  Speaker 7: that people will see if they use it now is
• 18:58
  Speaker 7: that it can do a task that takes a longer
• 19:00
  Speaker 7: time than they'd be able to delegate to a previous model.
• 19:05
  Speaker 7: And so I think the things that people should see
• 19:07
  Speaker 7: is just like as a normal American, if you want
• 19:09
  Speaker 7: a task, if you want to actually delegate a task
• 19:11
  Speaker 7: to it, it will be able to do that better
• 19:13
  Speaker 7: is one thing.
• 19:14
  Speaker 4: And then two is the thing that I think people.
• 19:16
  Speaker 7: Notice less, which is just that there's going to be
• 19:19
  Speaker 7: a huge amount of software being written by these models.
• 19:22
  Speaker 7: And so all of the products that people use are
• 19:25
  Speaker 7: going to be getting higher quality, cheaper over time, and
• 19:30
  Speaker 7: it'll have this effect of just like all of the
• 19:33
  Speaker 7: tools that we use now were written by people doing
• 19:37
  Speaker 7: engineering work, and now it's going to be written by
• 19:39
  Speaker 7: both people and super intelligences doing much, much more engineering
• 19:43
  Speaker 7: work than was possible before.
• 19:46
  Speaker 2: Absolutely, but I think it's doing so much more than this.
• 19:49
  Speaker 2: There's already multiple stories of AI, I'm sorry, superintelligence, superintelligence
• 19:56
  Speaker 2: finding a life-saving drug that can be used for a
• 20:01
  Speaker 2: disease that it wasn't understood that it.
• 20:03
  Speaker 3: Could safely treat.
• 20:04
  Speaker 2: So it's saving lives, I think it's having profound impacts
• 20:07
  Speaker 2: on education, but maybe each of you could talk For
• 20:10
  Speaker 2: ordinary Americans, what's super intelligence doing today and what's it
• 20:14
  Speaker 2: going to be able to do in a year?
• 20:16
  Speaker 3: I'm an ordinary American and I use.
• 20:22
  Speaker 5: Clyde to help me do research and instead of giving
• 20:27
  Speaker 5: me just answers in most of my research, maybe there's
• 20:30
  Speaker 5: comparative research that I'm doing and there's two or three
• 20:33
  Speaker 5: different approaches to something and I wanted to answer a
• 20:36
  Speaker 5: list of different questions and compare and contrast rates of
• 20:41
  Speaker 5: growth or rates of change.
• 20:43
  Speaker 3: Or whatever it is.
• 20:44
  Speaker 5: And so I describe literally every single question I have.
• 20:48
  Speaker 5: And nothing gives me more joy than, and I ask
• 20:51
  Speaker 5: it to create a nice report and plot it out.
• 20:54
  Speaker 3: Sometimes it's a pie chart, sometimes it's a bar graph,
• 20:56
  Speaker 3: and I just hit enter.
• 20:58
  Speaker 5: And nothing gives me more joy than just watching it go,
• 21:01
  Speaker 5: and then it comes back and it creates a beautiful report.
• 21:05
  Speaker 5: And then instead of researching all these different topics for four, six,
• 21:09
  Speaker 5: eight hours, which back in the old days I would
• 21:11
  Speaker 5: on a weekend, you know, learn something new, it goes
• 21:14
  Speaker 5: off and creates me this really, really beautiful final report
• 21:19
  Speaker 5: in about three minutes. And, you know, it's all condensed
• 21:23
  Speaker 5: and now you're getting smarter faster.
• 21:26
  Speaker 4: Love it. What about for you?
• 21:29
  Speaker 2: How are you seeing superintelligence helping ordinary Americans?
• 21:34
  Speaker 3: What do you think it'll do in a year or two?
• 21:36
  Speaker 3: Maybe a year or two. Take a year or two
• 21:38
  Speaker 3: from now.
• 21:39
  Speaker 6: Yeah, well, I think it's helping people organize their lives
• 21:42
  Speaker 6: and do better at their jobs. So I think it
• 21:46
  Speaker 6: really will become essential, if it isn't already, for people
• 21:50
  Speaker 6: to use AI in their professional lives, certainly, and in
• 21:55
  Speaker 6: their private lives.
• 21:56
  Speaker 4: And so AI is just an incredibly helpful, useful thing.
• 22:03
  Speaker 6: You know, I've heard many anecdotes of AI helping people
• 22:08
  Speaker 6: with medical issues. And in many cases where the doctor
• 22:12
  Speaker 6: was wrong and they submitted their x-rays or MRI to
• 22:18
  Speaker 6: AI and AI got it right.
• 22:22
  Speaker 4: But for AI, they would have died. And I've heard
• 22:24
  Speaker 4: many such cases. Awesome.
• 22:26
  Speaker 2: And Hopefully Opus 5-6 tells Tom, don't take that left,
• 22:32
  Speaker 2: take the right. I think it's important, these super intelligences,
• 22:39
  Speaker 2: they're imbued with the values of their creators. And we
• 22:43
  Speaker 2: do have, there are American values that we all share.
• 22:47
  Speaker 2: The First Amendment, free speech, commitment to truth. Can you
• 22:53
  Speaker 2: talk about what each of you are doing to imbue,
• 22:57
  Speaker 2: help make sure that these superintelligences are imbued with the
• 23:03
  Speaker 2: right values.
• 23:04
  Speaker 3: Are we allowed to say anything controversial? Say something controversial. Okay,
• 23:09
  Speaker 3: so the SIs are tools.
• 23:16
  Speaker 5: And the challenge with an SI with values, meaning it
• 23:20
  Speaker 5: has judgment, depending on how you use that SI, it
• 23:24
  Speaker 5: may judge you, and its effectiveness would be affected by
• 23:32
  Speaker 5: that judgment.
• 23:33
  Speaker 3: And so it's a little bit like a.
• 23:34
  Speaker 5: Steak knife, where if you were served medium well, it
• 23:40
  Speaker 5: cuts perfectly. But if you were served rare steak, and
• 23:44
  Speaker 5: it looks awfully like live meat, somehow it doesn't cut.
• 23:50
  Speaker 5: And so I think there's a question in my mind
• 23:55
  Speaker 5: yet about how much judgment you really want these tools
• 23:59
  Speaker 5: to have. And so the intentions of the user, obviously
• 24:06
  Speaker 5: I use the knife as an example, but you can
• 24:08
  Speaker 5: think of a whole bunch of other examples. The intentions
• 24:11
  Speaker 5: are a little bit different, whether you're a defender or
• 24:15
  Speaker 5: an attacker, for example, in cybersecurity. the defender's actions could
• 24:19
  Speaker 5: resemble an attacker. And you want the defender, you want
• 24:23
  Speaker 5: this SI to activate on behalf of the defender, even
• 24:29
  Speaker 5: though it appears to be an aggressive action.
• 24:33
  Speaker 4: Right.
• 24:33
  Speaker 2: I think we're running up on time, but maybe if
• 24:37
  Speaker 2: you guys could share, you spent a lot of time
• 24:39
  Speaker 2: at the White House. Is there anything that you can
• 24:42
  Speaker 2: share with the world about what came out of what
• 24:45
  Speaker 2: was evidently a long and productive lunch.
• 24:49
  Speaker 3: We did good.
• 24:49
  Speaker 4: Yeah.
• 24:51
  Speaker 6: We signed a joint declaration regarding SI safety. And I
• 24:59
  Speaker 6: think that will be published by the White House soon,
• 25:00
  Speaker 6: if it hasn't been already. But we agreed to a
• 25:05
  Speaker 6: number of things that include joint monitoring, board special committees,
• 25:10
  Speaker 6: and just generally grading each other's homework, which is a
• 25:14
  Speaker 6: lot better than if people just grade their own homework, uh,
• 25:17
  Speaker 6: for obvious reasons.
• 25:19
  Speaker 3: And it includes everything.
• 25:20
  Speaker 5: It includes everything from, um, rigorous internal controls, um, well-established methods,
• 25:27
  Speaker 5: well-established processes, um, very clearly documented intentions of the products
• 25:32
  Speaker 5: that you're building, internal controls to make sure that you're
• 25:35
  Speaker 5: delivering on those intentions, um, internal audit, um, that, that escalates, uh,
• 25:40
  Speaker 5: external audits, um, from, from a third party.
• 25:44
  Speaker 3: And, uh, And then industry sharing best practices.
• 25:48
  Speaker 5: I thought that the declaration has a lot of teeth.
• 25:54
  Speaker 3: I thought it was really well written.
• 25:55
  Speaker 5: And I think this is a practice that, quite frankly,
• 25:59
  Speaker 5: as standards for our industry going forward, every single company
• 26:03
  Speaker 5: ought to embrace. And so POTUS pulled it all together.
• 26:08
  Speaker 5: He watched over the meeting and said, was able to
• 26:12
  Speaker 5: bring everybody out. And, you know, I thought it was
• 26:15
  Speaker 5: a really well run.
• 26:16
  Speaker 4: Awesome.
• 26:17
  Speaker 2: Well, on behalf of all Americans, thank all of you
• 26:20
  Speaker 2: for doing everything you can to make the future as
• 26:22
  Speaker 2: awesome and optimistic for American Americans as you can.
• 26:26
  Speaker 3: And you are. So thank you.
• 26:33
  Speaker 1: Thanks for listening.
• 26:34
  Speaker 4: See you in the next episode.