【スターゲートがAIの利用爆増を支える】オープンAI・アルトマンCEO／トランプ氏は直感で良い判断を下す／１時間で１００万ユーザー増「ジブリ風」画像生成で浮上した課題／発端はサプライチェーンの限界

00:17:42
https://www.youtube.com/watch?v=Khk1KkqExxw

Resumo

TLDRこのインタビューでは、AIの急速な進化とそれに伴う計算能力の必要性について、OpenAIのCEOが語っています。特に、Stargateプロジェクトの立ち上げや、AIモデルの需要の急増が強調され、5000億ドルの資金が必要な理由が説明されています。また、AIが雇用に与える影響や、科学的発見を加速させる可能性についても言及されています。技術の進化が新たな雇用を生む一方で、既存の職業が消失するリスクについても触れられています。

Conclusões

  • 🚀 AIの需要が急増している
  • 💰 5000億ドルの資金が必要
  • 🔍 AIは科学的発見を加速する
  • 🤖 AIは新たな雇用を生む可能性がある
  • ⚙️ Stargateプロジェクトは大規模な計算インフラを目指す
  • 📈 技術の進化により効率的なチップが開発される
  • 🌍 AIの競争は激化している
  • 📊 計算能力の不足が課題
  • 💡 AIは人類の理解を深める
  • 🔗 Deepseekの効率的な電力供給方法に注目

Linha do tempo

  • 00:00:00 - 00:05:00

    このセクションでは、AIモデルの需要が予想以上に高まっていることが強調され、Stargateプロジェクトの必要性が説明されています。特に、GPT-4のリリース後、計算能力の限界に直面し、より多くのコンピューティングリソースと接続性が求められるようになったことが述べられています。

  • 00:05:00 - 00:10:00

    次に、Stargateプロジェクトの資金調達とその背後にあるパートナーシップについての詳細が語られています。特に、MicrosoftやSoftBankとの協力が重要であり、5000億ドルの資金が必要な理由が説明されています。AIの需要が急増している中で、OpenAIが持続可能なビジネスモデルを確立できるかどうかについての自信も示されています。

  • 00:10:00 - 00:17:42

    最後に、AIの進化がもたらす雇用の変化についての懸念が取り上げられています。AIは多くの仕事を奪う一方で、新しい仕事を生み出す可能性があることが強調され、技術の進化がもたらす影響についての認識が求められています。特に、AIが科学的発見を加速させる可能性についての期待が語られています。

Mapa mental

Vídeo de perguntas e respostas

  • Stargateプロジェクトとは何ですか？

    Stargateプロジェクトは、AIモデルの需要に応えるための大規模な計算インフラストラクチャーの構築を目指しています。

  • なぜ5000億ドルの資金が必要なのですか？

    これは、今後数年間の成長予測に基づいて必要な計算能力をカバーするための資金です。

  • AIはどのように雇用に影響を与えるのですか？

    AIは既存の職業を消失させる一方で、新たな職業を生む可能性があります。

  • OpenAIはどのようにして持続可能な企業になるのですか？

    現在の成長と収益性の見込みに基づいて、OpenAIは持続可能な企業になると考えています。

  • Stargateはどのように設計されていますか？

    データセンターの設計は、計算需要に基づいて最適化されています。

  • AIの未来についてどう考えていますか？

    AIは科学的発見を加速させ、人類の理解を深めると信じています。

  • Deepseekの効率的な電力供給方法についてどう思いますか？

    Deepseekのチームは優れた成果を上げていますが、私たちの方法も効率的です。

  • AIの需要はどのように変化していますか？

    AIの需要は急増しており、計算能力が不足しています。

  • AIの進化はどのように進むと考えていますか？

    技術の進化により、より効率的なチップやアルゴリズムが開発されるでしょう。

  • AIの競争についてどう考えていますか？

    私たちは日々改善を目指しており、競争に対して最善を尽くしています。

Ver mais resumos de vídeos

Obtenha acesso instantâneo a resumos gratuitos de vídeos do YouTube com tecnologia de IA!
Legendas
en
Rolagem automática:
  • 00:00:00
    [Music]
  • 00:00:03
    Hello. Good to see you. Thank you.
  • 00:00:07
    So, I saw what's happening in Abalene
  • 00:00:09
    with my own eyes. Isn't it cool? It's
  • 00:00:11
    happening. Take me back to the beginning
  • 00:00:12
    before Stargate. What did you start
  • 00:00:14
    seeing that made you realize we need
  • 00:00:16
    more compute, more power, more
  • 00:00:18
    connectivity that you were hitting a
  • 00:00:19
    limit and needed to scale?
  • 00:00:22
    We used to think a lot about the compute
  • 00:00:23
    we would need to train the models. And
  • 00:00:27
    what we didn't used to think about as
  • 00:00:29
    much is how much people were going to
  • 00:00:30
    use these models. Um I mean we did think
  • 00:00:32
    about it. It just turned out people want
  • 00:00:33
    to use the models much more than we
  • 00:00:36
    imagined. And a couple of years ago
  • 00:00:39
    maybe after the launch of GPT4 in Chhat
  • 00:00:42
    GBT it really started to hit like oh man
  • 00:00:45
    this is this is like gone from a lot of
  • 00:00:47
    compute to like the biggest
  • 00:00:48
    infrastructure project in history. Yeah.
  • 00:00:50
    And we started thinking about ways to do
  • 00:00:51
    it and we started trying to understand
  • 00:00:52
    where the limits in the supply chain
  • 00:00:54
    were and
  • 00:00:57
    and out of that emerged Stargate. Um
  • 00:00:59
    there were other smaller things that we
  • 00:01:01
    did first. We have you know worked with
  • 00:01:04
    uh other partners. We've worked with
  • 00:01:05
    Microsoft to build out like very
  • 00:01:07
    gigantic amounts of compute. But um you
  • 00:01:10
    know this is like the next step in that
  • 00:01:11
    evolution, right? I mean you traveled
  • 00:01:13
    around the world talking to people about
  • 00:01:15
    this. How did you get to talking with
  • 00:01:17
    Masayoshi's son and Larry Ellison and
  • 00:01:19
    how did you all come together? Um I'm
  • 00:01:23
    well I've known Masa for a long time. Um
  • 00:01:25
    but I I was on a in 203
  • 00:01:31
    I did sort of two long trips. I think
  • 00:01:32
    that was the year that's all it's all
  • 00:01:34
    about. Um I I did two kind of long trips
  • 00:01:37
    around the world and a lot of it was to
  • 00:01:39
    talk to developers and governments. Uh,
  • 00:01:42
    but a lot of it was also to just really
  • 00:01:44
    try to like get my head around the
  • 00:01:46
    supply chain. This is not something I've
  • 00:01:48
    done before. Like I had not before
  • 00:01:49
    thought about like what it was going to
  • 00:01:50
    take to get compute up and running at
  • 00:01:52
    this scale. And there are a lot of like
  • 00:01:54
    hard pieces. Um, and on one of those
  • 00:01:58
    trips, uh, I met with Masa. Um, Masa has
  • 00:02:02
    like thought for a long time about, uh,
  • 00:02:05
    chip fabs in particular, but really kind
  • 00:02:07
    of the whole thing that it takes. and we
  • 00:02:09
    got to speaking about what it would um
  • 00:02:11
    you know what it would take to do
  • 00:02:13
    compute at this kind of scale and then
  • 00:02:15
    it took us a while to to figure out I
  • 00:02:16
    mean it is a is a complex supply chain
  • 00:02:19
    with a lot of partners and obviously a
  • 00:02:20
    lot of capital and you know Soft Bank's
  • 00:02:22
    obviously your financial partner Oracle
  • 00:02:23
    is the technical partner why couldn't
  • 00:02:25
    you get what you needed from Microsoft I
  • 00:02:28
    mean we do get a lot of great stuff from
  • 00:02:29
    Microsoft but I think this is more than
  • 00:02:31
    any one company can deliver mic we have
  • 00:02:33
    Microsoft will do a lot of compute for
  • 00:02:35
    us a lot we're very happy about that why
  • 00:02:38
    call Stargate. Um, it began as a code
  • 00:02:42
    name and sometimes code names stick,
  • 00:02:45
    but it means something, right? Oh. Um,
  • 00:02:48
    the design of one of the very early
  • 00:02:52
    layouts of a data center looked a little
  • 00:02:53
    bit like Stargate from the from the
  • 00:02:55
    show. Mhm. A futuristic wormhole. Not
  • 00:02:58
    like that at all. Just Okay. You
  • 00:03:00
    announced this at the White House the
  • 00:03:02
    day after the inauguration.
  • 00:03:04
    How did this get to President Trump's
  • 00:03:05
    desk and and what was that moment like?
  • 00:03:09
    Well, the president is super interested
  • 00:03:10
    in infrastructure and he's made uh you
  • 00:03:13
    know a big priority which I think is
  • 00:03:15
    wonderful about uh permitting the energy
  • 00:03:20
    production and the facilities for the
  • 00:03:22
    data centers. Uh so that was kind of how
  • 00:03:23
    the conversation started. Okay. We're in
  • 00:03:26
    the middle of a step change for AI and
  • 00:03:28
    clearly what we saw just in the last
  • 00:03:29
    week from you alone. um you're seeing
  • 00:03:32
    what's next on the road map. How does
  • 00:03:34
    that inform what the design of the data
  • 00:03:38
    centers of the future need to be?
  • 00:03:43
    I
  • 00:03:44
    I mean I think there's like a lot of
  • 00:03:46
    details we're learning about how to
  • 00:03:47
    design these data centers, what you want
  • 00:03:50
    um really at all levels from like the
  • 00:03:52
    chip to the sort of architecture of the
  • 00:03:54
    whole data center.
  • 00:03:56
    But but the main the main thing that's
  • 00:03:58
    been on my mind and I think on many
  • 00:04:00
    people's mind is just how much inference
  • 00:04:02
    demand there is. Um we are crazily
  • 00:04:06
    constrained. We have a gigantic comput
  • 00:04:09
    fleet like gigantic gigantic and yet
  • 00:04:12
    still um if we had twice as much we
  • 00:04:14
    would be able to offer much better
  • 00:04:15
    products and services and there's just
  • 00:04:16
    no shortage of demand. So
  • 00:04:20
    for me, there are all the technical
  • 00:04:22
    lessons about what we've learned and how
  • 00:04:25
    we want to build this, but mostly we
  • 00:04:26
    just want a lot. Explain the math to us.
  • 00:04:29
    How does it all add up to needing $500
  • 00:04:31
    billion? Because folks out there don't
  • 00:04:33
    get the accounting, who's putting in
  • 00:04:36
    what, how much funding is really
  • 00:04:38
    secured. Um, how does it add up in terms
  • 00:04:40
    of like what do we need that for? How
  • 00:04:42
    how does the funding come together? Like
  • 00:04:43
    explain that number 500 billion. Oh. Um,
  • 00:04:47
    well, that covers the capacity we think
  • 00:04:49
    we need for the next few years given our
  • 00:04:50
    growth projections. Uh, you know, we we
  • 00:04:54
    the the interesting question is if we
  • 00:04:56
    had if we knew how to get a trillion
  • 00:04:57
    dollars right now, which we don't. Um,
  • 00:05:00
    would we be able to deploy that
  • 00:05:02
    profitably in the next few years?
  • 00:05:05
    And I'm not sure about that, but I feel
  • 00:05:07
    confident we can like make 500 billion
  • 00:05:08
    of value back. Let's talk about that
  • 00:05:10
    because you just raised even more money,
  • 00:05:12
    you know, generally for right into Open
  • 00:05:15
    AI. Um, how confident it's a lot of
  • 00:05:17
    money, tens of billions of dollars. How
  • 00:05:19
    confident are you that Open AI is going
  • 00:05:22
    to be a financially sustainable and
  • 00:05:24
    profitable enough company to justify all
  • 00:05:26
    of this investment? It's looking like
  • 00:05:29
    we're doing really well. I mean, like we
  • 00:05:31
    have to we are definitely doing
  • 00:05:33
    something unprecedented. Uh, but
  • 00:05:36
    you know, it seems like a I like I feel
  • 00:05:40
    confident in the bet. Doesn't mean
  • 00:05:42
    something can't go wrong. You You
  • 00:05:44
    tweeted that the GPUs were melting. You
  • 00:05:48
    Yeah, I didn't realize people were going
  • 00:05:48
    to take that literally. I I I mean, I
  • 00:05:51
    get that it was a joke, but it's very
  • 00:05:53
    hot, but like the metal is not actually
  • 00:05:55
    melting. GPU shortage, bro. Um, that the
  • 00:05:59
    team can't sleep. You're so busy. What
  • 00:06:01
    happens when a launch goes viral and how
  • 00:06:04
    does Stargate solve for this? Well,
  • 00:06:08
    first of all, this level of viral this
  • 00:06:09
    is an unusual thing. This last week, I
  • 00:06:11
    don't think this has happened in the
  • 00:06:12
    history of tech to any company before.
  • 00:06:14
    Like the I don't know of any I've seen
  • 00:06:18
    viral moments, but I have never seen
  • 00:06:19
    anyone have to deal with an influx of
  • 00:06:21
    usage like this. You added a million
  • 00:06:22
    users in an hour or something like that.
  • 00:06:24
    Yeah, I mean more some hours, but yeah.
  • 00:06:27
    um it was like a it was like un
  • 00:06:29
    unprecedented wild and also like making
  • 00:06:32
    an image is not a it's not exactly like
  • 00:06:34
    a low compute task the way we do it with
  • 00:06:36
    the new image then so um we had to do a
  • 00:06:39
    lot of very unnatural things we had to
  • 00:06:40
    you know borrow compute capacity from
  • 00:06:42
    research we had to slow down some other
  • 00:06:43
    features because it's not like we have
  • 00:06:45
    like hundreds of thousands of GPUs
  • 00:06:47
    sitting around just like spinning idly
  • 00:06:49
    so with if we had more GPUs we would be
  • 00:06:53
    more able to
  • 00:06:57
    handle demand surges like this and also
  • 00:06:59
    we wouldn't have to put such
  • 00:07:00
    restrictions on. Um I was just this
  • 00:07:03
    morning making the list of what we have
  • 00:07:05
    what we'd like to launch in the next few
  • 00:07:08
    months and I was like you know I had the
  • 00:07:11
    list over here and I had like my rough
  • 00:07:12
    math of when we're getting GPUs here and
  • 00:07:14
    where I thought we could get the
  • 00:07:15
    efficiency gains and I was like this is
  • 00:07:16
    not going to work like we're going to
  • 00:07:19
    have to like pull some of these things
  • 00:07:20
    make some trade-offs limit some things I
  • 00:07:22
    don't like but this is not all of this
  • 00:07:23
    is not going to fit here. So it's that
  • 00:07:26
    directly linked. Yeah. I mean you can
  • 00:07:28
    find efficiency gains and we tried to
  • 00:07:30
    do. You can limit features and usage and
  • 00:07:31
    we do that too. But uh you know more
  • 00:07:33
    compute means we can give you more AI
  • 00:07:35
    and you can use it for images or writing
  • 00:07:37
    software whatever you want. Elon's got
  • 00:07:39
    his own data centers and XAI just bought
  • 00:07:42
    X which means he's got access to all of
  • 00:07:44
    that additional data. He claims Grock is
  • 00:07:47
    the smartest AI on Earth. How do you
  • 00:07:49
    think it compares? I don't use it much.
  • 00:07:52
    I think it's probably a very good model.
  • 00:07:54
    Mhm. I mean obviously there are lots of
  • 00:07:55
    competitors out there. There's a lot of
  • 00:07:57
    good models. Uh
  • 00:08:03
    I think I think good models
  • 00:08:06
    will become very bountiful, very
  • 00:08:07
    plentiful. Um so what's your edge going
  • 00:08:09
    to be? The best infrastructure layer and
  • 00:08:12
    the best top of the stack. You know we
  • 00:08:14
    have way more people use chatbt than use
  • 00:08:17
    any other uh AI service. Way more. Um,
  • 00:08:21
    and I think you'll see with a lot of the
  • 00:08:23
    new features we roll out that that will
  • 00:08:25
    become um, more of an advantage over
  • 00:08:27
    time. So once this is all built, what's
  • 00:08:29
    the grand vision?
  • 00:08:32
    Give people tools to let them do
  • 00:08:34
    whatever they're going to do better. Um,
  • 00:08:36
    you know, it's it's this is in some
  • 00:08:37
    sense this is like it is different this
  • 00:08:39
    time, but in many other senses, this is
  • 00:08:41
    just like another tool and another piece
  • 00:08:43
    of technological history. people will
  • 00:08:46
    use it and unleash their creative energy
  • 00:08:48
    and make all sorts of stuff that you and
  • 00:08:50
    I love or some stuff that we don't. Um
  • 00:08:52
    or stuff to just entertain themselves.
  • 00:08:56
    Personally, the area I'm most excited
  • 00:08:58
    about is AI for scientific discovery. Um
  • 00:09:00
    I think that will, you know, we're not
  • 00:09:02
    there yet, but we're not far away. I
  • 00:09:04
    think 2025 will be a world where we have
  • 00:09:07
    agents do a lot of work, but work of the
  • 00:09:10
    kind of work and things we already know
  • 00:09:11
    how to do. I'm hopeful that 2026 will be
  • 00:09:13
    a big year of like really uh new
  • 00:09:17
    scientific progress. Do you envision
  • 00:09:19
    multiple Stargates on every continent?
  • 00:09:21
    Like where else are we going to see you
  • 00:09:23
    break ground? You will see them on other
  • 00:09:24
    continents. Yes. Everywhere. I don't
  • 00:09:27
    know about everywhere. Like this is this
  • 00:09:29
    is like you you went to visit one of
  • 00:09:31
    this like these are hard things to do.
  • 00:09:33
    Well, look, this is an audacious bet on
  • 00:09:35
    the future. What are the risks?
  • 00:09:38
    Um
  • 00:09:40
    I mean maybe you know people stop
  • 00:09:42
    wanting to like pay for AI services and
  • 00:09:44
    then we have a difficult financial
  • 00:09:45
    position.
  • 00:09:48
    How do you feel about one chipmaker
  • 00:09:50
    Nvidia having so much power over the
  • 00:09:52
    industry the future of the industry they
  • 00:09:54
    make an incredible product and if you
  • 00:09:57
    make an incredible product and people
  • 00:09:58
    want it this is what happens to you. I
  • 00:10:01
    do want to talk about the jobs thing
  • 00:10:02
    because obviously we saw so many people
  • 00:10:03
    working there. There is this lofty
  • 00:10:05
    promise that AI data centers are going
  • 00:10:07
    to create thousands and thousands of
  • 00:10:08
    jobs. Meanwhile, AI is destroying jobs
  • 00:10:11
    elsewhere. And I feel like even you know
  • 00:10:13
    there is serious anxiety out there.
  • 00:10:16
    People are scared. Totally. Even among
  • 00:10:18
    the best engineers and technologists,
  • 00:10:20
    people are scared. Um what do you say to
  • 00:10:23
    those people?
  • 00:10:24
    AI is for sure going to change a lot of
  • 00:10:28
    jobs. Totally take some jobs away.
  • 00:10:30
    Create a bunch of new ones. Mhm. This is
  • 00:10:32
    like the kind of this is what happens
  • 00:10:34
    with technology. In fact, if I think if
  • 00:10:36
    you look at the history of
  • 00:10:39
    the world like
  • 00:10:41
    technological driven job change or
  • 00:10:44
    whatever you call it when like one class
  • 00:10:46
    of jobs go away and another one pop up
  • 00:10:47
    like that's very consistent. It happens
  • 00:10:49
    at a it's punctuated but like that's
  • 00:10:51
    just been happening for a long time. And
  • 00:10:54
    the thing that is different this time is
  • 00:10:55
    just the rate with which it looks like
  • 00:10:56
    it will happen. Mhm. Um, the thing I
  • 00:10:59
    think the world is not ready for, like
  • 00:11:00
    people have maybe abstractly thought
  • 00:11:02
    like, okay, it's going to be a better
  • 00:11:03
    programmer than me. It's going to be,
  • 00:11:04
    you know, better at customer support and
  • 00:11:06
    whatever. Um, I don't think the world
  • 00:11:08
    has really had the humanoid robots
  • 00:11:10
    moment yet. And I don't think that's
  • 00:11:12
    very far away from like a visceral like,
  • 00:11:14
    oh man, this is going to do a lot of
  • 00:11:16
    things that people used to do. So, so
  • 00:11:17
    yeah, it's coming. And we we have always
  • 00:11:19
    tried to just be super honest about what
  • 00:11:21
    we think the impact may be, realizing
  • 00:11:23
    that we'll be wrong on a lot of details.
  • 00:11:25
    What happens when the humanoid robots
  • 00:11:26
    get here?
  • 00:11:28
    Um,
  • 00:11:31
    I mean they'll obviously do a lot of
  • 00:11:32
    jobs, but what I the point I was trying
  • 00:11:33
    to make is I think like the first day
  • 00:11:35
    you're like walking down the street and
  • 00:11:37
    there's like seven robots that walk past
  • 00:11:38
    you doing things or whatever, it's going
  • 00:11:40
    to feel very sci-fi. Mhm. Deepseek
  • 00:11:43
    appears to have found a more efficient
  • 00:11:44
    way to power AI. Was that a moment of
  • 00:11:47
    rethink for you? And are you doing
  • 00:11:49
    anything differently now? Um, I think
  • 00:11:52
    the Deep Seek team is very talented and
  • 00:11:53
    did a lot of good things. I don't think
  • 00:11:55
    they figured out something like way more
  • 00:11:56
    efficient than what we figured out. But
  • 00:11:58
    do you think there is a more efficient
  • 00:12:00
    way to build AI? I'm we will we have
  • 00:12:03
    made incredible efficiency strides year
  • 00:12:05
    over year and I'm sure we'll keep doing
  • 00:12:07
    that in the future. So if that's the
  • 00:12:08
    case, why are you building all this?
  • 00:12:12
    If we had an AI that we could offer at
  • 00:12:14
    onetenth of the price of current AI, um
  • 00:12:17
    I think people would use it 20 times as
  • 00:12:19
    much and we would still need twice as
  • 00:12:21
    much compute to satisfy the then current
  • 00:12:23
    demand. So let's talk about that Jevans
  • 00:12:26
    paradox like how do you think Jevans
  • 00:12:27
    paradox applies here that you know
  • 00:12:30
    technological progress means
  • 00:12:31
    paradoxically you're you're going to be
  • 00:12:33
    using more resources doesn't
  • 00:12:37
    it doesn't this never seemed like a
  • 00:12:38
    paradox to me at all um like you know we
  • 00:12:42
    talk about supply demand curves and
  • 00:12:43
    elasticity and all sorts of other things
  • 00:12:45
    and then people are like ah but I'm
  • 00:12:46
    going to trick you with Jeb's paradox
  • 00:12:47
    and it's like this is just the way the
  • 00:12:49
    world works most of the time but you do
  • 00:12:51
    think there will be more efficient ways
  • 00:12:53
    what What are those? Is it better chips?
  • 00:12:54
    Is it is it all of the above? It will we
  • 00:12:57
    will have better chips. We will have
  • 00:12:58
    better energy sources. We will have
  • 00:12:59
    better algorithms. Um just we will
  • 00:13:03
    optimize. I mean this is just this is
  • 00:13:04
    like what industry does. We will
  • 00:13:06
    optimize everything. Mhm. There's no
  • 00:13:08
    question that China is going to be a
  • 00:13:09
    formidable player in AI. How do you
  • 00:13:12
    think about OpenAI's chances to win a
  • 00:13:15
    global race?
  • 00:13:19
    You know, probably I should have some
  • 00:13:21
    deep thoughts about that.
  • 00:13:23
    We're doing the best we can. Like we
  • 00:13:25
    just wake up every day thinking how we
  • 00:13:26
    can be a little bit better. Uh
  • 00:13:28
    I don't know how to like I don't know
  • 00:13:29
    how our actions would change based off
  • 00:13:31
    of like some deep answer to that
  • 00:13:33
    question. President Trump, you know, you
  • 00:13:35
    said he cares about infrastructure.
  • 00:13:36
    President Trump is in power at a pivotal
  • 00:13:39
    time for AI development. What do you
  • 00:13:41
    think his mark on this moment will be? I
  • 00:13:43
    think he will get to make some of the
  • 00:13:45
    most important decisions anyone in the
  • 00:13:46
    world has gotten to make related to AI.
  • 00:13:50
    Um, and you know, I'm optimistic he's
  • 00:13:54
    really
  • 00:13:55
    he'll really do the right thing there,
  • 00:13:56
    but I I don't like he has an unenviable
  • 00:13:59
    job, right? The velocity of in we were
  • 00:14:03
    talking about the launches. The velocity
  • 00:14:04
    of innovation here is just mind-blowing.
  • 00:14:07
    The speed at which you're releasing new
  • 00:14:09
    things, the speed of what's coming up,
  • 00:14:11
    what I saw rising from the red dirt in
  • 00:14:13
    Abolene. How do you personally grapple
  • 00:14:16
    with the the pace of it all? That's one
  • 00:14:19
    thing that's really impressed me about
  • 00:14:20
    President Trump, by the way, is his
  • 00:14:22
    ability to just like
  • 00:14:26
    understand the whole industry and all
  • 00:14:28
    the changes and quickly seem to have
  • 00:14:30
    very good intuition and make good
  • 00:14:31
    decisions about it. Uh, while things are
  • 00:14:34
    changing so fast has has really been
  • 00:14:36
    quite impressive to me. Um,
  • 00:14:39
    but what about you personally? Yeah. I
  • 00:14:40
    was going to say I
  • 00:14:44
    I'm in it every day and and so I don't
  • 00:14:48
    like
  • 00:14:50
    it's like watching your own kid grow
  • 00:14:51
    like you day to day you just see every
  • 00:14:53
    change and so it's like not as striking.
  • 00:14:56
    Um it does feel like it's going very
  • 00:14:58
    fast. It's certainly
  • 00:15:03
    I certainly think if I could like
  • 00:15:04
    transport myself back three years ago,
  • 00:15:07
    it would seem like unimaginable
  • 00:15:09
    progress.
  • 00:15:10
    But, you know, daytoday you can kind of
  • 00:15:12
    get used to anything. Well, since you
  • 00:15:14
    mentioned kids, you just had a baby. It
  • 00:15:16
    did. Has that reframed anything for you?
  • 00:15:20
    T. I mean, like, look, I don't think I
  • 00:15:21
    have anything nonclée to say here, but
  • 00:15:23
    it is the best most amazing thing ever
  • 00:15:24
    and it totally rewired all of my
  • 00:15:26
    priorities. I remember in like the first
  • 00:15:28
    hour I like felt this neurochemical
  • 00:15:30
    change and it happened so fast. I was
  • 00:15:31
    like, "Oh, I get to like observe this.
  • 00:15:33
    Like I am being like neurochemically
  • 00:15:35
    hacked, but I'm like noticing it
  • 00:15:37
    happening. I'm totally fine with this.
  • 00:15:38
    This is working. is great, but like
  • 00:15:40
    everything is going to be different now.
  • 00:15:41
    What about how everything that you're
  • 00:15:43
    doing here and building all this what it
  • 00:15:46
    means for humanity? Has it reframed any
  • 00:15:48
    of that? Um,
  • 00:15:53
    a lot of people have said, "I'm very
  • 00:15:54
    happy you're having a kid because I
  • 00:15:56
    think you'll make better I think you'll
  • 00:15:58
    make like better decisions to whatever
  • 00:15:59
    degree you have." Like a you got to like
  • 00:16:01
    make decisions here for humanity as a
  • 00:16:03
    whole.
  • 00:16:05
    I really wanted to get it right before
  • 00:16:07
    I'd do the best I could. I I still
  • 00:16:09
    really want to now. It uh
  • 00:16:16
    it somehow Yes, it somehow does feel
  • 00:16:18
    like it's different, but I can't
  • 00:16:19
    articulate how. As you keep moving
  • 00:16:22
    forward, if Chat GPT were to discover a
  • 00:16:25
    Stargate, a futuristic wormhole that we
  • 00:16:28
    could all travel through,
  • 00:16:30
    what's on the other side? Like, what do
  • 00:16:32
    you see that the rest of us don't?
  • 00:16:35
    If Chip could like transport us to the
  • 00:16:37
    future. Yeah. Through a Stargate. A
  • 00:16:39
    wormhole. Please humor me on the
  • 00:16:40
    wormhole. I have no idea. I mean, no,
  • 00:16:43
    no, no one knows, right? Like, it's
  • 00:16:44
    going to
  • 00:16:48
    it's going to discover like AI. I I
  • 00:16:50
    believe AI will help dramatically, like
  • 00:16:53
    unbelievably dramatically accelerate
  • 00:16:57
    the pace of science and human discovery
  • 00:16:59
    and our understanding of what's
  • 00:17:00
    possible.
  • 00:17:02
    But like, do I think I could have sat
  • 00:17:04
    here in 1905 and told you what we're
  • 00:17:08
    about to discover in physics
  • 00:17:10
    and that 40 years later we would like
  • 00:17:12
    have an atomic bomb? Definitely not. Uh
  • 00:17:16
    I and I I think I am way too self-aware
  • 00:17:19
    of my own limitations to sit here and
  • 00:17:21
    try to say I can like tell you what's on
  • 00:17:23
    the other side of that wormhole. I have
  • 00:17:24
    no idea. But net good. Yes. happy uh
  • 00:17:28
    better a better a better world up and
  • 00:17:31
    down not better in every way but yeah I
  • 00:17:33
    think up to the right yeah up to the
  • 00:17:35
    right with some choppiness Yes.
Etiquetas
  • AI
  • Stargate
  • 計算能力
  • OpenAI
  • 雇用
  • 科学的発見
  • 技術の進化
  • Deepseek
  • 資金調達
  • 競争