[00:01.560 -> 00:07.640] Only four companies in the world are worth more than $2 trillion – Microsoft, Apple,
[00:07.640 -> 00:12.880] Alphabet, parent company of Google, and computer chip maker Nvidia.
[00:12.880 -> 00:19.400] The California-based company saw its stock market value soar from $1 trillion to $2 trillion
[00:19.400 -> 00:26.560] in just eight months this past year, fueled by the insatiable demand for its cutting-edge technology, the
[00:26.560 -> 00:31.760] hardware and software that make today's artificial intelligence possible.
[00:31.760 -> 00:37.840] We wondered how a company founded in 1993 to improve video game graphics turned into
[00:37.840 -> 00:41.360] a titan of 21st century AI.
[00:41.360 -> 00:53.420] So we went to Silicon Valley to meet Nvidia's 61-year-old co-founder and CEO, Jensen Huang, who has no doubt AI is about to change everything.
[00:54.960 -> 00:58.000] The story will continue in a moment.
[01:13.160 -> 01:14.460] At NVIDIA's annual developers conference this past March, the mood wasn't just upbeat.
[01:16.680 -> 01:17.400] It was downright giddy.
[01:30.520 -> 01:31.620] More than 11,000 enthusiasts, software developers, tech moguls, and happy shareholders filed into San Jose's pro hockey arena to kick off a four-day AI extravaganza.
[01:37.360 -> 01:37.880] They came to see this man, Jensen Huang, CEO of NVIDIA.
[01:39.980 -> 01:40.640] Welcome to GTC!
[01:44.600 -> 01:45.100] What was that like for you to walk out on that stage and see that?
[01:47.540 -> 01:48.040] You know, Bill, I'm an engineer, not a performer.
[01:53.440 -> 01:53.600] When I walked out there and all of the people going crazy, it took the breath out of me.
[01:56.640 -> 01:58.000] And so I was the scariest I've ever been. I'm still scared.
[01:59.400 -> 01:59.820] You'd never know it.
[02:06.180 -> 02:06.280] Clad in his signature cool black outfit, Jensen shared the stage with NVIDIA-powered robots.
[02:07.320 -> 02:07.620] Let me finish up real quick.
[02:10.380 -> 02:10.600] And shared his vision of an AI future.
[02:12.320 -> 02:18.900] A new industrial revolution. It reminded us of the transformational moment when Apple's Steve Jobs unveiled the iPhone.
[02:19.460 -> 02:24.940] Jensen Huang unveiled NVIDIA's latest graphics processing unit, or GPU.
[02:25.440 -> 02:26.520] This is Blackwell.
[02:26.880 -> 02:31.400] Designed in America but made in Taiwan like most advanced semiconductors,
[02:31.920 -> 02:35.080] Blackwell, he says, is the fastest chip ever.
[02:35.460 -> 02:37.360] Google is gearing up for Blackwell.
[02:37.360 -> 02:40.520] The whole industry is gearing up for Blackwell.
[02:41.020 -> 02:47.540] NVIDIA ushered in the AI revolution with its game-changing GPU, a single chip able
[02:47.540 -> 02:53.860] to process a myriad of calculations all at once, not sequentially like more standard chips.
[02:54.480 -> 03:02.080] The GPU is the engine of NVIDIA's AI computer, enabling it to rapidly absorb a fire hose of
[03:02.080 -> 03:06.340] information. It does quadrillions of calculations a second.
[03:06.340 -> 03:08.500] It's just insane numbers.
[03:08.500 -> 03:10.880] Is it doing things now that surprise you?
[03:10.880 -> 03:12.720] We're hoping that it does things that surprise us.
[03:12.720 -> 03:14.020] That's the whole point.
[03:14.020 -> 03:16.220] In some areas, like drug discovery,
[03:16.220 -> 03:19.860] designing better materials that are lighter, stronger.
[03:19.860 -> 03:23.860] We need artificial intelligence to help us explore the universe
[03:23.860 -> 03:26.240] in places that we could have never done ourselves.
[03:26.400 -> 03:28.540] Let me show you. Here, Bill, look at this.
[03:28.760 -> 03:36.920] Jensen took us around the GTC convention hall to show us what AI has made possible in just the past few years.
[03:37.060 -> 03:38.340] I'm making your drink now.
[03:38.480 -> 03:40.660] Some creations were dazzling.
[03:40.660 -> 03:43.460] This is a digital twin of the Earth.
[03:46.020 -> 03:53.660] dazzling. This is a digital twin of the earth. Once it learns how to calculate weather, it can calculate and predict weather 3,000 times faster than a supercomputer and a thousand times less
[03:53.660 -> 04:03.480] energy. But NVIDIA's AI revolution extends far beyond this hall. Blue metallic spaceship.
[04:03.480 -> 04:05.080] And let's generate something.
[04:06.540 -> 04:08.400] Pinar Seyhan Demirda is originally from Istanbul,
[04:08.840 -> 04:11.300] but co-founded Qubrick near Boston.
[04:11.980 -> 04:13.320] Her AI application
[04:13.320 -> 04:15.420] uses NVIDIA's GPUs
[04:15.420 -> 04:16.740] to instantly turn
[04:16.740 -> 04:18.220] a simple text prompt
[04:18.220 -> 04:20.040] into a virtual movie set
[04:20.040 -> 04:21.820] for a fraction of the cost
[04:21.820 -> 04:23.360] of today's backdrops.
[04:23.360 -> 04:24.360] This isn't something
[04:24.360 -> 04:26.120] that's already planned.
[04:26.260 -> 04:28.960] No, we're doing it in real time. It's live.
[04:29.160 -> 04:30.820] Is Hollywood knocking at your door?
[04:31.700 -> 04:33.300] We're getting a lot of love.
[04:34.880 -> 04:37.200] Nearby at Generate Biomedicines,
[04:37.200 -> 04:40.540] Dr. Alex Snyder, head of research and development,
[04:40.880 -> 04:46.680] is using NVIDIA's technology to create protein-based drugs. She was surprised
[04:46.680 -> 04:49.480] at first to see they showed promise in the lab.
[04:49.820 -> 04:54.260] Initially, when I was told about the application of AI to drug development, I sort of rolled
[04:54.260 -> 04:59.140] my eyes and said, yeah, you know, show me the data. And then I looked at the data, and
[04:59.140 -> 05:00.080] it was very compelling.
[05:01.140 -> 05:07.720] Dr. Snyder's team asks its AI models to create new proteins to fight specific diseases
[05:07.720 -> 05:09.800] like cancer and asthma.
[05:09.800 -> 05:13.940] A new way to defeat the coronavirus is now in clinical trials.
[05:13.940 -> 05:19.940] You're now working with proteins that do not exist in nature, that you're coming
[05:19.940 -> 05:22.440] up with by way of AI?
[05:22.440 -> 05:23.520] Yes.
[05:23.520 -> 05:26.020] We are actually generating what we call de novo,
[05:26.020 -> 05:30.060] completely new structures that have not existed before.
[05:30.060 -> 05:31.520] Do you trust it?
[05:31.520 -> 05:34.320] As scientists, we can't trust, we have to test.
[05:34.320 -> 05:36.380] We're not putting Franken-signs into people.
[05:36.380 -> 05:37.920] We're taking what's known,
[05:37.920 -> 05:39.860] and we're really pushing the field,
[05:39.860 -> 05:42.240] we're pushing the biology to make drugs
[05:42.240 -> 05:45.240] that look like regular drugs drugs but function even better.
[05:45.900 -> 05:48.200] This is a technology that will only get better from here.
[05:48.560 -> 05:54.520] Brett Adcock is CEO of Figure, a Silicon Valley startup with funding from NVIDIA.
[05:55.060 -> 06:02.520] Look at his answer to labor shortages, an NVIDIA GPU-driven prototype called Figure One.
[06:03.260 -> 06:07.240] I think what's been really extraordinary is the pace of progress we've made in 21 months.
[06:07.240 -> 06:08.540] From zero to this in 21 months.
[06:08.540 -> 06:09.540] Zero to this, yeah.
[06:09.540 -> 06:12.340] We were walking this robot in under a year
[06:12.340 -> 06:14.240] since I incorporated the company.
[06:14.240 -> 06:17.040] Could you do this without NVIDIA's technology?
[06:17.040 -> 06:20.260] We think they're arguably the best in the world at this.
[06:20.260 -> 06:23.000] I don't know if this would be possible without them.
[06:23.000 -> 06:26.020] I'm here to assist with tasks as requested.
[06:26.640 -> 06:32.020] We were amazed that Figure 1 is not just walking, but seemed to reason.
[06:32.780 -> 06:34.640] Hand me something healthy.
[06:35.660 -> 06:36.020] On it.
[06:36.440 -> 06:41.940] Figure 1 was able to understand I wanted the orange, not the packaged snack.
[06:41.940 -> 06:42.800] Thank you.
[06:43.580 -> 06:44.920] It's not yet perfected.
[06:44.980 -> 06:45.000] You're going to get it. not the packaged snack. Thank you. It's not yet perfected.
[06:45.000 -> 06:46.000] You're going to get it.
[06:46.000 -> 06:48.000] But the early results are so promising,
[06:48.000 -> 06:53.000] German automaker BMW plans to start testing the robot
[06:53.000 -> 06:56.000] in its South Carolina factory this year.
[06:56.000 -> 06:59.000] I think there's an opportunity to ship billions of robots
[06:59.000 -> 07:02.000] in the coming decades onto the planet.
[07:02.000 -> 07:04.000] Billions.
[07:04.000 -> 07:08.700] I would think that a lot of workers would look at that as,
[07:08.700 -> 07:11.360] this robot is taking my job.
[07:11.360 -> 07:13.520] I think over time, AI and robotics
[07:13.520 -> 07:17.940] will start doing more and more of what humans can and better.
[07:17.940 -> 07:20.120] But what about the worker?
[07:20.120 -> 07:22.360] The workers work for companies.
[07:22.360 -> 07:27.060] And so companies, when they become more productive, earnings increase.
[07:27.680 -> 07:33.160] I've never seen one company that had earnings increase and not hire more people.
[07:33.160 -> 07:37.160] There are some jobs that are going to become obsolete.
[07:38.040 -> 07:39.340] Well, let me offer it this way.
[07:39.620 -> 07:43.880] I believe that you still want human in the loop because we have good judgment,
[07:44.340 -> 07:46.000] because there are circumstances that the machines
[07:46.000 -> 07:48.000] are just not going to understand.
[07:48.000 -> 07:50.000] The futuristic NVIDIA campus
[07:50.000 -> 07:54.000] sits just down the road from its modest birthplace,
[07:54.000 -> 07:56.000] this Denny's in San Jose.
[07:56.000 -> 07:58.000] Good morning.
[07:58.000 -> 08:01.000] Where 31 years ago, NVIDIA was just an idea.
[08:01.000 -> 08:03.000] My goodness.
[08:03.000 -> 08:06.920] When he was 15, Jensen Huang worked as a dishwasher at Denny's.
[08:07.320 -> 08:14.100] As a 30-year-old electrical engineer married with two children, he and two friends, NVIDIA
[08:14.100 -> 08:21.560] co-founders Chris Malachowski and Curtis Preem, envisioned a whole new way of processing video
[08:21.560 -> 08:22.500] game graphics.
[08:22.780 -> 08:25.760] So we came here, right here to this denny's, sat right back there,
[08:26.120 -> 08:28.600] and the three of us decided to start the company.
[08:29.360 -> 08:31.000] Frankly, I had no idea how to do it.
[08:32.060 -> 08:32.820] And nor did they.
[08:33.080 -> 08:34.400] None of us knew how to do anything.
[08:35.280 -> 08:36.280] Their big idea?
[08:37.120 -> 08:39.520] Accelerate the processing power of computers
[08:39.520 -> 08:41.620] with a new graphics chip.
[08:42.260 -> 08:44.220] Their initial attempt flopped
[08:44.220 -> 08:48.420] and nearly bankrupted the company in 1996.
[08:48.420 -> 08:53.980] And the genius of the engineers and Chris and Curtis, we pivoted to the right way of
[08:53.980 -> 08:55.300] doing things.
[08:55.300 -> 08:58.320] And created their groundbreaking GPU.
[08:58.320 -> 09:07.160] The chip took video games from this to this today. Completely changed computer graphics,
[09:07.680 -> 09:08.420] saved the company,
[09:09.100 -> 09:11.500] launched us into the stratosphere.
[09:12.160 -> 09:14.220] Just eight years after Denny's,
[09:14.480 -> 09:17.440] NVIDIA earned a spot in the S&P 500.
[09:18.180 -> 09:19.740] Jensen then set his sights
[09:19.740 -> 09:21.900] on developing the software and hardware
[09:21.900 -> 09:25.580] for a revolutionary GPU-driven supercomputer,
[09:26.060 -> 09:29.320] which would take the company far beyond video games.
[09:29.800 -> 09:32.520] To Wall Street, it was a risky bet.
[09:32.960 -> 09:36.300] To early developers of AI, it was a revelation.
[09:37.020 -> 09:38.880] Was that luck or was that vision?
[09:39.440 -> 09:41.980] That was luck founded by vision.
[09:42.160 -> 09:43.980] We invented this capability.
[09:44.980 -> 09:46.380] And then one day the
[09:46.380 -> 09:52.000] researchers that were creating deep learning discovered this architecture
[09:52.000 -> 09:55.600] because this architecture turns out to have been perfect for them.
[09:55.600 -> 09:56.800] Perfect for AI.
[09:56.800 -> 09:57.800] Perfect for AI.
[09:57.800 -> 09:59.800] This is the first one we've ever shipped.
[09:59.800 -> 10:14.520] In 2016 Jensen delivered NVIDIA's AI supercomputer, the first of its kind, to Elon Musk, then a board member of OpenAI, which used it to create the building blocks of ChatGPT.
[10:14.880 -> 10:15.320] How are you?
[10:15.680 -> 10:20.260] When AI took off, so did Jensen Huang's reputation.
[10:21.620 -> 10:22.780] Can we get a picture?
[10:23.020 -> 10:23.520] Yeah, yeah.
[10:23.520 -> 10:26.000] He's now a Silicon Valley celebrity.
[10:26.000 -> 10:33.000] He told us the boy who immigrated from Taiwan at age nine could never have conceived of this.
[10:33.000 -> 10:41.000] It is the most extraordinary thing, Bill, that a normal dishwasher busboy could grow up to be this.
[10:41.000 -> 10:47.460] There's no magic. It's just 61 years of hard work every single day.
[10:48.120 -> 10:53.640] I don't think there's anything more than that. We met a humble Jensen at Denny's. Back at
[10:53.640 -> 11:00.600] NVIDIA's headquarters in Santa Clara, we saw he can be intense. Let me tell you what some of the
[11:00.600 -> 11:05.480] people who you work with said about you. Demanding. Perfectionist.
[11:05.920 -> 11:07.220] Not easy to work for.
[11:07.900 -> 11:08.740] All that sound right?
[11:09.020 -> 11:09.760] Perfectly, yeah.
[11:10.740 -> 11:12.000] It should be like that.
[11:12.520 -> 11:14.980] If you want to do extraordinary things,
[11:15.740 -> 11:16.640] it shouldn't be easy.
[11:17.200 -> 11:18.480] All right, guys, keep up the good work.
[11:18.900 -> 11:20.840] NVIDIA has never done better.
[11:21.440 -> 11:22.780] Investors are bullish.
[11:22.780 -> 11:27.060] But last year, more than 600 top AI scientists,
[11:27.260 -> 11:34.160] ethicists, and others signed this statement urging caution, warning of AI's risk to humanity.
[11:34.760 -> 11:38.000] When I talk to you and I hear you speak, part of me goes,
[11:38.600 -> 11:43.980] gee whiz. And the other part of me goes, oh my God, what are we in for?
[11:44.200 -> 11:45.180] Yeah, yeah.
[11:45.260 -> 11:45.980] Which one is it?
[11:46.220 -> 11:47.900] It's both. It's both.
[11:48.400 -> 11:50.560] Yeah, you're feeling all the right feelings. I feel both.
[11:51.220 -> 11:51.820] You feel both?
[11:51.840 -> 11:52.940] Sure, sure.
[11:53.400 -> 12:01.360] Humanity will have the choice to see themselves inferior to machines or superior to machines.
[12:01.360 -> 12:08.400] Pinar Seyhan Demirda is an AI optimist, though she named her company Kubrick, an homage to
[12:08.400 -> 12:12.300] Stanley Kubrick, the director of 2001, A Space Odyssey.
[12:12.300 -> 12:14.440] Hello, Hal, do you read me?
[12:14.440 -> 12:18.960] In that film, Hal, the AI computer, goes rogue.
[12:18.960 -> 12:22.100] Open the pod bay doors, Hal.
[12:22.100 -> 12:23.800] I'm sorry, Dave.
[12:23.800 -> 12:25.400] I'm afraid I can't do that.
[12:25.400 -> 12:33.240] I think that's what worries people about AI, that we will lose control of it.
[12:33.740 -> 12:42.320] Just because a machine can do faster calculations, comparisons, and analytical solution creation, that doesn't make you smarter than you.
[12:42.720 -> 12:44.380] It simply computates faster.
[12:42.340 -> 12:42.800] that doesn't make you smarter than you.
[12:44.380 -> 12:45.380] It simply computates faster.
[12:47.280 -> 12:47.880] In my world, in my belief,
[12:50.360 -> 12:50.760] smarts have to do with your capacity to love,
[12:53.020 -> 12:54.340] create, expand, transcend.
[12:58.140 -> 12:58.680] These are qualities that no machine can ever bear,
[13:00.360 -> 13:00.740] that are reserved to only humans.
[13:02.360 -> 13:02.860] There is something going on.
[13:09.800 -> 13:10.460] Jensen Huang sees an AI future of progress and prosperity, not one with machines as our masters.
[13:13.000 -> 13:13.380] We can only hope he's right.
[13:15.960 -> None] Thank you all for coming. Thank you.
沒有留言:
張貼留言