1 00:00:01,520 --> 00:00:02,240 Speaker 1: From our heart. 2 00:00:02,360 --> 00:00:06,800 Speaker 2: We're Innovation, Money and Power Collie in Silicon Valley, NBN. 3 00:00:07,120 --> 00:00:11,640 Speaker 2: This is Bloomberg Technology with Caroline Hyde and Ed Ludlow. 4 00:00:27,400 --> 00:00:29,920 Speaker 3: Live from sunny San Francisco and Caroline Hyde. 5 00:00:29,720 --> 00:00:30,520 Speaker 4: And our Ed Ludlow. 6 00:00:30,600 --> 00:00:34,519 Speaker 5: This is a special edition of Bloomberg Technology. Coming up 7 00:00:34,640 --> 00:00:37,760 Speaker 5: will bring you live coverage from our technology event as 8 00:00:37,760 --> 00:00:41,080 Speaker 5: we speak with the CEOs and visionaries that are driving 9 00:00:41,200 --> 00:00:43,360 Speaker 5: change in Silicon Valley and beyond. 10 00:00:43,520 --> 00:00:46,599 Speaker 3: Now at this hour, we're speaking with the CEOs of Arm, 11 00:00:46,880 --> 00:00:50,240 Speaker 3: of Hugging, Face, of Writer, AI and more as part 12 00:00:50,360 --> 00:00:51,680 Speaker 3: of our live special. 13 00:00:51,720 --> 00:00:52,880 Speaker 4: And a lot more to come. 14 00:00:52,880 --> 00:00:55,480 Speaker 5: We'll push your head to our keynote speakers later today. 15 00:00:55,760 --> 00:00:59,880 Speaker 5: That includes Adam Newman, Whitney Wolf Heard, Evan Spiegel and more. 16 00:01:00,080 --> 00:01:00,720 Speaker 4: And while it. 17 00:01:00,640 --> 00:01:02,440 Speaker 5: May not be a surprise to anyone too, did it's 18 00:01:02,440 --> 00:01:05,840 Speaker 5: certainly not a surprised to us. Artificial intelligence is probably 19 00:01:05,880 --> 00:01:10,560 Speaker 5: the overriding theme of the event. It's AI everything, even 20 00:01:10,600 --> 00:01:12,440 Speaker 5: in some cases if you're not an AI company. 21 00:01:12,640 --> 00:01:14,119 Speaker 3: Yeah, I mean I think it has to be. Even 22 00:01:14,160 --> 00:01:16,080 Speaker 3: if you're not the AI company at heart, you're thinking 23 00:01:16,080 --> 00:01:17,560 Speaker 3: about how you adapt to it, push it forward and 24 00:01:17,560 --> 00:01:19,760 Speaker 3: bring out the productivity. But Ultimately, how do we live 25 00:01:19,840 --> 00:01:22,520 Speaker 3: up to the hype? The valuations are so extraordinary in 26 00:01:22,560 --> 00:01:24,880 Speaker 3: the private markets, they've been pretty heavy in the public 27 00:01:24,920 --> 00:01:27,160 Speaker 3: markets as well. Are we really seeing the level of 28 00:01:27,160 --> 00:01:29,640 Speaker 3: producted divity and growth and real use cases? 29 00:01:29,680 --> 00:01:33,440 Speaker 5: Do mindication that these events are always very interesting, very engaging. 30 00:01:33,560 --> 00:01:35,960 Speaker 5: Everyone's very positive, But I would say in the background, 31 00:01:36,319 --> 00:01:39,800 Speaker 5: there's a talent war, frankly, and people are running out 32 00:01:39,800 --> 00:01:40,480 Speaker 5: of cash. 33 00:01:40,280 --> 00:01:42,200 Speaker 4: So we've got to ask those difficult questions too. 34 00:01:42,319 --> 00:01:45,839 Speaker 3: And geopolitics, how are you navigating the issue of China 35 00:01:45,920 --> 00:01:48,520 Speaker 3: as well? So so much to talk about in here 36 00:01:48,560 --> 00:01:52,920 Speaker 3: and the now. We're talking about one key infrastructure playing 37 00:01:53,000 --> 00:01:55,760 Speaker 3: When it comes to artificial intelligence, we're of course going 38 00:01:55,800 --> 00:01:58,560 Speaker 3: to be talking about the chip design firm ARM, which 39 00:01:58,800 --> 00:02:01,480 Speaker 3: has really bounced offlows in terms of its share price 40 00:02:01,520 --> 00:02:03,960 Speaker 3: throughout the trading of today. We're currently over the last 41 00:02:03,960 --> 00:02:05,960 Speaker 3: two days down by one point few percent. Coming out 42 00:02:05,960 --> 00:02:09,040 Speaker 3: after the bell with its earnings, tapid forecast head was 43 00:02:09,080 --> 00:02:11,320 Speaker 3: what the market seemed to be focusing in on, even 44 00:02:11,320 --> 00:02:13,840 Speaker 3: though they absolutely smashed it in terms of their fiscal 45 00:02:13,880 --> 00:02:16,760 Speaker 3: fourth quarter numbers and their fourth first quarter. To look ahead, 46 00:02:17,120 --> 00:02:19,720 Speaker 3: let's stick in to some of that caution with Renee 47 00:02:19,760 --> 00:02:22,600 Speaker 3: has is the arm CEO. Renee wonderful to have time 48 00:02:22,639 --> 00:02:24,720 Speaker 3: with you. And look, there does seem to be a 49 00:02:24,760 --> 00:02:28,880 Speaker 3: worry about your full year forecasts? Are you being cautious? 50 00:02:30,480 --> 00:02:33,160 Speaker 2: Well, thanks both for having me Ed and Caroline. We 51 00:02:33,320 --> 00:02:36,120 Speaker 2: just came off a record year in terms of revenue. 52 00:02:36,120 --> 00:02:38,200 Speaker 2: We were up twenty percent a little bit over twenty 53 00:02:38,280 --> 00:02:43,120 Speaker 2: percent our fiscal twenty three from twenty twenty two, and 54 00:02:43,160 --> 00:02:46,320 Speaker 2: we're actually forecasting even higher growth this year, north of 55 00:02:46,400 --> 00:02:49,919 Speaker 2: twenty percent. And we also signaled to the markets yesterday 56 00:02:49,960 --> 00:02:53,799 Speaker 2: that in twenty five, twenty six, twenty seven, we see 57 00:02:53,800 --> 00:02:57,880 Speaker 2: that growth continuing. So we have incredible visibility to our 58 00:02:57,919 --> 00:03:01,160 Speaker 2: business and we're very, very confident of growth rate going forward. 59 00:03:04,440 --> 00:03:07,240 Speaker 5: We're just seeing your shares actually ticking to positive territory 60 00:03:07,240 --> 00:03:11,240 Speaker 5: rene up now six tenths of one percent. The underlying 61 00:03:11,320 --> 00:03:14,720 Speaker 5: story is the build out in AI infrastructure. Right, we're 62 00:03:14,720 --> 00:03:20,680 Speaker 5: talking about data center powered GP by GPUs. Your numbers 63 00:03:20,680 --> 00:03:24,320 Speaker 5: were good. Tell me about the underlying demand then about 64 00:03:24,320 --> 00:03:26,920 Speaker 5: the long term and the addressable market you think is 65 00:03:26,960 --> 00:03:28,320 Speaker 5: either intact or is not. 66 00:03:30,040 --> 00:03:30,200 Speaker 3: Well. 67 00:03:30,240 --> 00:03:34,000 Speaker 2: I think this AI buildout, as you describe, or maybe 68 00:03:34,040 --> 00:03:38,360 Speaker 2: said another way, just expanding capacity to run these foundation models, 69 00:03:38,400 --> 00:03:39,840 Speaker 2: to do more and more training, to do more and 70 00:03:39,920 --> 00:03:40,480 Speaker 2: more inference. 71 00:03:41,440 --> 00:03:42,680 Speaker 3: We really are only. 72 00:03:42,520 --> 00:03:44,800 Speaker 2: At the very beginning because when you start to think 73 00:03:44,840 --> 00:03:49,000 Speaker 2: about the capabilities that this could unleash, whether it's around healthcare, 74 00:03:49,400 --> 00:03:54,920 Speaker 2: farmer research, productivity, gains, call centers, we're still in the 75 00:03:55,000 --> 00:03:58,080 Speaker 2: very very early days. That all starts with having to 76 00:03:58,120 --> 00:04:00,320 Speaker 2: do this level of training and inprints in the cloud, 77 00:04:00,720 --> 00:04:03,920 Speaker 2: but it ultimately will find itself in every single edge device, 78 00:04:04,080 --> 00:04:09,120 Speaker 2: whether that's a PC, your smartphone, your car, and whether 79 00:04:09,160 --> 00:04:11,880 Speaker 2: it's all those devices I've mentioned from the data center 80 00:04:11,920 --> 00:04:14,440 Speaker 2: to the edge devices. They all run on ARM. So 81 00:04:14,680 --> 00:04:17,760 Speaker 2: we have incredible visibility to where this is all going, 82 00:04:17,760 --> 00:04:19,680 Speaker 2: which is why we're very confident in the growth rates. 83 00:04:20,279 --> 00:04:22,320 Speaker 2: They're also one of the big problems you've got with 84 00:04:22,520 --> 00:04:25,960 Speaker 2: all of these AI data centers is around energy and power. 85 00:04:26,400 --> 00:04:28,800 Speaker 2: So power efficiency being so key, it's what ARM is 86 00:04:28,880 --> 00:04:32,600 Speaker 2: really good at. Increasingly we're seeing the most complex applications 87 00:04:32,680 --> 00:04:35,680 Speaker 2: moving to ARM and most sophisticated training ship on the 88 00:04:35,680 --> 00:04:39,719 Speaker 2: planet that was just announced Grace Blackwell, Well that's based 89 00:04:39,720 --> 00:04:44,160 Speaker 2: on ARM. 90 00:04:44,320 --> 00:04:46,400 Speaker 3: Okay, so you're managing to really think that you're going 91 00:04:46,440 --> 00:04:50,120 Speaker 3: to be the server play as well as the PC play, 92 00:04:50,120 --> 00:04:51,760 Speaker 3: the cell phone play, and I want to focus in 93 00:04:51,800 --> 00:04:53,760 Speaker 3: on the cell phone play, Renee, because that's been where 94 00:04:53,760 --> 00:04:56,480 Speaker 3: your bread and butter has been in history. How are 95 00:04:56,480 --> 00:04:59,320 Speaker 3: we looking from a smartphone perspective? Is the market looking 96 00:04:59,360 --> 00:05:01,960 Speaker 3: strong to you? We've had many a mixed message coming 97 00:05:01,960 --> 00:05:03,320 Speaker 3: from China to MOD for example. 98 00:05:05,200 --> 00:05:08,000 Speaker 2: Overall, what we've seen the smartphone market briftly for ARM 99 00:05:08,080 --> 00:05:10,880 Speaker 2: has been quite a good growth rate in terms of royalties. 100 00:05:11,080 --> 00:05:14,320 Speaker 2: Our version nine which is now being used in many 101 00:05:14,360 --> 00:05:17,599 Speaker 2: of the premium mobile phones, that drives a higher royalty 102 00:05:17,680 --> 00:05:21,039 Speaker 2: rate for ARM. There's also more complex CPUs that go 103 00:05:21,120 --> 00:05:24,839 Speaker 2: into that that's also better for ARM and going forward carrolling. 104 00:05:24,920 --> 00:05:26,680 Speaker 2: One of the things that we're seeing, and it's not 105 00:05:26,760 --> 00:05:30,520 Speaker 2: just in smartphones, is that as these AI models are 106 00:05:30,520 --> 00:05:34,600 Speaker 2: moving so fast, the hardware can't keep up with the software. 107 00:05:35,000 --> 00:05:38,919 Speaker 2: The software innovation is happening so quickly that by the 108 00:05:38,960 --> 00:05:41,680 Speaker 2: time the hardware is ready to run those models, everyone 109 00:05:41,720 --> 00:05:44,680 Speaker 2: wishes they had more performance, they had more efficiency. 110 00:05:45,160 --> 00:05:46,280 Speaker 4: So what does that mean for ARM? 111 00:05:46,760 --> 00:05:50,239 Speaker 2: It's driving growth in our licensing activity. People are looking 112 00:05:50,240 --> 00:05:53,440 Speaker 2: to do more and more design ships faster and faster, 113 00:05:53,920 --> 00:05:56,160 Speaker 2: and that's all good for us going forward. So I 114 00:05:56,200 --> 00:05:58,040 Speaker 2: think going forward you're going to see more and more 115 00:05:58,040 --> 00:06:01,080 Speaker 2: innovation happening, not only in the smartphones, across all these 116 00:06:01,120 --> 00:06:01,880 Speaker 2: edge devices. 117 00:06:05,320 --> 00:06:07,520 Speaker 3: What's interesting in AY is it's hard to keep up 118 00:06:07,560 --> 00:06:11,000 Speaker 3: with the pace of geopolitical change as well. The latest 119 00:06:11,000 --> 00:06:13,479 Speaker 3: news coming that Huawei, of course is not going to 120 00:06:13,480 --> 00:06:16,400 Speaker 3: have access to poll Kong to Intel chips. You were, 121 00:06:16,400 --> 00:06:19,800 Speaker 3: of course a UK based company, but are affected by 122 00:06:19,920 --> 00:06:23,720 Speaker 3: US policies. Has this impacted your business? The limitations of 123 00:06:23,760 --> 00:06:28,120 Speaker 3: Huawei's access to chip designed to chip technology to licenses. 124 00:06:29,800 --> 00:06:32,960 Speaker 2: Yeah, So that issue they referred to specifically was when 125 00:06:33,040 --> 00:06:35,960 Speaker 2: Huawei was placed on the entity list I think twenty nineteen, 126 00:06:36,040 --> 00:06:39,839 Speaker 2: twenty twenty, companies had to apply for a license to 127 00:06:40,080 --> 00:06:42,960 Speaker 2: exempt them to ship to Huawei. So a number of 128 00:06:43,040 --> 00:06:47,120 Speaker 2: companies asked for those licenses, they got those licenses. Now 129 00:06:47,120 --> 00:06:49,919 Speaker 2: those licenses are being revoked. We don't follow in that 130 00:06:50,040 --> 00:06:52,599 Speaker 2: category in any way, shape or form. We didn't apply 131 00:06:52,720 --> 00:06:55,440 Speaker 2: for any licenses at the time to share. We complied 132 00:06:55,480 --> 00:06:57,880 Speaker 2: with the export controls as they were laid out. So 133 00:06:58,080 --> 00:07:00,920 Speaker 2: there's really a non event for us in terms of 134 00:07:00,960 --> 00:07:02,719 Speaker 2: what you're seeing with Qualcoman or Intel. 135 00:07:06,160 --> 00:07:09,119 Speaker 5: We are speaking live to the ARMS CEO, Renee has 136 00:07:09,200 --> 00:07:12,520 Speaker 5: We're on the ground here at Bloomberg Tech in San Francisco. 137 00:07:12,880 --> 00:07:15,880 Speaker 5: Last week, Renee Christiano mom was on the show telling 138 00:07:15,920 --> 00:07:19,160 Speaker 5: Caroline and I, this is the year of the AIPC. 139 00:07:20,000 --> 00:07:22,280 Speaker 5: You were asked about that on your earning school last 140 00:07:22,360 --> 00:07:25,560 Speaker 5: night and you gave a slightly different answer. And maybe 141 00:07:25,560 --> 00:07:28,800 Speaker 5: it's not the year of the AIPC more the twelve 142 00:07:28,800 --> 00:07:31,480 Speaker 5: to thirty six month window. And you don't want to 143 00:07:31,520 --> 00:07:34,080 Speaker 5: see just one PC supplier, you said you'd like to 144 00:07:34,120 --> 00:07:38,680 Speaker 5: see two or three. What's your beef with Qualcom? 145 00:07:38,800 --> 00:07:38,880 Speaker 1: Now? 146 00:07:38,960 --> 00:07:43,040 Speaker 2: When I look at the PC ecosystem, one large ecosystem 147 00:07:43,080 --> 00:07:45,000 Speaker 2: has already moved to ARM in a very big way. 148 00:07:46,000 --> 00:07:49,320 Speaker 2: Apple is now one based on ARM. All the Apple 149 00:07:49,360 --> 00:07:52,680 Speaker 2: silicon is based on ARM. And you see amazingly good 150 00:07:52,760 --> 00:07:57,440 Speaker 2: products relative to what they've delivered, fantastic battery life, performance, 151 00:07:57,520 --> 00:08:00,720 Speaker 2: thin and light, no fans. When you think about the 152 00:08:00,760 --> 00:08:03,760 Speaker 2: Windows market, it's a very different market. It's highly fragmented. 153 00:08:03,800 --> 00:08:06,800 Speaker 2: You have lots of different players. The ecosystem matters, the 154 00:08:06,920 --> 00:08:11,880 Speaker 2: channel matters, price points matter, high end gaming machines versus 155 00:08:12,840 --> 00:08:16,480 Speaker 2: low end devices that are like cloud enabled. So what 156 00:08:16,520 --> 00:08:20,280 Speaker 2: does all that mean. It generally has meant that breath 157 00:08:20,840 --> 00:08:24,640 Speaker 2: vendor choice, multiple options to provide a full scope is 158 00:08:24,640 --> 00:08:28,000 Speaker 2: what matters. And what I'm hearing is over the next 159 00:08:28,000 --> 00:08:30,080 Speaker 2: couple of years, the Windows ecosystem is going to be 160 00:08:30,080 --> 00:08:33,000 Speaker 2: able to afford that. And I think over the next 161 00:08:33,320 --> 00:08:37,280 Speaker 2: two three years, I do believe Windows Unarmed will be real. 162 00:08:37,559 --> 00:08:41,000 Speaker 2: I think you'll see multiple players, multiple price points, multiple units, 163 00:08:41,320 --> 00:08:43,280 Speaker 2: and I think you'll see meaningful market share that we 164 00:08:43,320 --> 00:08:46,120 Speaker 2: start to gain the kind of performance you see in 165 00:08:46,160 --> 00:08:48,080 Speaker 2: the other ecosystem. I think we'll find its way into 166 00:08:48,120 --> 00:08:52,640 Speaker 2: the Windows ecosystem. 167 00:08:52,920 --> 00:08:54,800 Speaker 4: Rennie, I wanted to talk about geography really quick. 168 00:08:55,000 --> 00:08:57,360 Speaker 5: We're here in San Francisco, right there's a lot about 169 00:08:57,679 --> 00:09:01,520 Speaker 5: Americas are in d focus on AI related chips. 170 00:09:01,800 --> 00:09:05,240 Speaker 4: Are you seeing this sort of equivalent activity in. 171 00:09:05,120 --> 00:09:08,480 Speaker 5: Europe, for example, any of your customers outside of those markets. 172 00:09:09,640 --> 00:09:11,400 Speaker 2: Yeah, well I'm in San Francisco today too, so I 173 00:09:11,400 --> 00:09:14,760 Speaker 2: will see you a little bit later. But in general, 174 00:09:15,040 --> 00:09:19,240 Speaker 2: I think the geopolitics are something that all tech CEOs 175 00:09:19,240 --> 00:09:23,400 Speaker 2: are now having to figure out and work with AI models, 176 00:09:23,520 --> 00:09:28,200 Speaker 2: foundation models, sovereign clouds, thinking about what level of training 177 00:09:28,240 --> 00:09:32,720 Speaker 2: takes place in a country, versus outside the country where 178 00:09:32,720 --> 00:09:34,679 Speaker 2: the weights sit, et cetera. That these are all the 179 00:09:34,800 --> 00:09:37,720 Speaker 2: kind of things that politicians have never really had to 180 00:09:37,760 --> 00:09:40,080 Speaker 2: think about in the past. So we're involved in a 181 00:09:40,080 --> 00:09:43,800 Speaker 2: lot of those conversations, whether that's in the United States, 182 00:09:43,840 --> 00:09:46,480 Speaker 2: whether that's in Europe, and really just trying to understand 183 00:09:46,480 --> 00:09:49,960 Speaker 2: it because any lawmakers in all these jurisdictions are just 184 00:09:50,000 --> 00:09:52,440 Speaker 2: trying to figure it all out. And as I mentioned before, 185 00:09:52,880 --> 00:09:56,040 Speaker 2: as the software and models are moving so fast, it's 186 00:09:56,080 --> 00:09:58,720 Speaker 2: difficult for everyone to keep up. But we are central 187 00:09:58,760 --> 00:09:59,880 Speaker 2: to all those discussions. 188 00:10:03,559 --> 00:10:06,160 Speaker 3: Renee, what's been keeping up is your valuation? 189 00:10:06,679 --> 00:10:06,959 Speaker 4: Boy? 190 00:10:07,520 --> 00:10:10,320 Speaker 3: I mean, do you think there's too much exuberance around 191 00:10:10,400 --> 00:10:13,800 Speaker 3: AI valuations out there? Are you going to make the 192 00:10:13,880 --> 00:10:15,800 Speaker 3: most of it? By well, we talked to one point 193 00:10:15,800 --> 00:10:17,080 Speaker 3: of listing in the UK too. 194 00:10:19,000 --> 00:10:21,320 Speaker 2: Yeah, you know, I don't think about the valuations as 195 00:10:21,400 --> 00:10:23,839 Speaker 2: much as I just think about the AI opportunity, which 196 00:10:23,960 --> 00:10:28,600 Speaker 2: I frankly believe is undercalled in terms of just what 197 00:10:28,679 --> 00:10:31,199 Speaker 2: it's going to mean relative to society and what it 198 00:10:31,240 --> 00:10:34,240 Speaker 2: can do for our planet. I think again we are 199 00:10:34,280 --> 00:10:37,679 Speaker 2: in very very early days in terms of the capabilities 200 00:10:37,720 --> 00:10:41,800 Speaker 2: of what this can unleash for our society incredibly excited 201 00:10:41,840 --> 00:10:43,760 Speaker 2: to be part of it. But I don't think we're 202 00:10:43,800 --> 00:10:45,680 Speaker 2: part of a hype cycle at all. I think there's 203 00:10:45,720 --> 00:10:48,679 Speaker 2: a lot of innovation taking place, and you know, frankly, 204 00:10:48,920 --> 00:10:52,160 Speaker 2: the innovation that's taking place, any inventions that we're seeing, 205 00:10:52,600 --> 00:10:55,400 Speaker 2: it's just breathtaking. So no, I don't personally view it 206 00:10:55,400 --> 00:10:56,439 Speaker 2: as a hype cycle at all. 207 00:11:00,080 --> 00:11:03,000 Speaker 5: They has I'm CEO really grateful you actually be here 208 00:11:03,000 --> 00:11:06,080 Speaker 5: with us later today on site at Bloomberg Tech. Your 209 00:11:06,120 --> 00:11:08,959 Speaker 5: stock open pretty low, and I think it's just a 210 00:11:09,000 --> 00:11:11,000 Speaker 5: little bit higher now during the conversation we've had. 211 00:11:11,040 --> 00:11:11,760 Speaker 4: Thank you so much. 212 00:11:11,800 --> 00:11:13,719 Speaker 5: All Right, coming up on the program, we're going to 213 00:11:13,760 --> 00:11:17,120 Speaker 5: be joined by Clem DeLong, CEO of Hugging Face. That's 214 00:11:17,120 --> 00:11:19,520 Speaker 5: coming up next. Stay tuned, we'll be right back. Is 215 00:11:19,600 --> 00:11:35,360 Speaker 5: Bloomberg Technology. Welcome back to this special edition of Bloomberg 216 00:11:35,400 --> 00:11:38,559 Speaker 5: Technology live in San Francisco at the Bloomberg Tech event 217 00:11:38,679 --> 00:11:40,840 Speaker 5: and artificial intelligence Surprise. 218 00:11:40,920 --> 00:11:43,080 Speaker 4: Surprise is sort of the overarching theme. 219 00:11:43,400 --> 00:11:45,319 Speaker 5: We've got a pretty good guest to talk about that 220 00:11:45,360 --> 00:11:48,640 Speaker 5: with and discuss all things large language model with Clemed Along, 221 00:11:48,960 --> 00:11:52,679 Speaker 5: CEO of Hugging Face. You made this prediction which we're 222 00:11:52,720 --> 00:11:55,880 Speaker 5: going to hold you to account on that by the 223 00:11:55,960 --> 00:11:58,160 Speaker 5: end of this calendar year, and I appreciate we're not 224 00:11:58,200 --> 00:12:03,880 Speaker 5: even halfway. Source models would be equivalent to the best 225 00:12:04,000 --> 00:12:07,560 Speaker 5: closed source models. Give us a status check of that 226 00:12:07,679 --> 00:12:08,520 Speaker 5: prediction place. 227 00:12:08,880 --> 00:12:10,000 Speaker 4: I think it's already happened. 228 00:12:10,480 --> 00:12:13,560 Speaker 6: Open source now is better than closed source for most 229 00:12:13,840 --> 00:12:19,960 Speaker 6: use cases. We've specialized customized models on the companies like 230 00:12:20,040 --> 00:12:24,160 Speaker 6: data sets. I have my meta reband glasses here that 231 00:12:24,240 --> 00:12:25,720 Speaker 6: are powered by Lammatri. 232 00:12:26,320 --> 00:12:26,480 Speaker 4: Right. 233 00:12:26,559 --> 00:12:30,240 Speaker 6: We've seen so many now use cases being powered by 234 00:12:30,280 --> 00:12:34,079 Speaker 6: open source models, and most of the big tech companies 235 00:12:34,240 --> 00:12:37,439 Speaker 6: are now publishing open models. Just last month, we've seen 236 00:12:38,240 --> 00:12:42,120 Speaker 6: Apple releasing open models on hogging Face. We've seen Nvidia, 237 00:12:42,280 --> 00:12:46,600 Speaker 6: we've seen Snowflakes, We've seen Data Bricks, we've seen Microsoft. 238 00:12:46,760 --> 00:12:48,640 Speaker 6: All of them now are publishing open models. 239 00:12:48,679 --> 00:12:51,040 Speaker 5: There maybe some people in attendance who don't agree with him, 240 00:12:51,040 --> 00:12:52,240 Speaker 5: which is why I asked the question. 241 00:12:52,400 --> 00:12:55,080 Speaker 3: Well, also Microsoft published one and then quickly with DURI 242 00:12:55,480 --> 00:12:58,719 Speaker 3: some of the reporting because they hadn't stress tested the 243 00:12:58,840 --> 00:13:01,880 Speaker 3: large language model enough. I hadn't whittled out some of 244 00:13:01,880 --> 00:13:05,800 Speaker 3: the toxicity checks. In particular, how are you feeling about 245 00:13:06,000 --> 00:13:09,160 Speaker 3: the way in which large language models are growing and 246 00:13:09,280 --> 00:13:11,719 Speaker 3: the way in which governance is developing around it. 247 00:13:12,120 --> 00:13:14,240 Speaker 6: Well, we're starting to see that the most important question 248 00:13:14,440 --> 00:13:17,880 Speaker 6: is concentration of power. Right for such an important technology, 249 00:13:17,880 --> 00:13:20,920 Speaker 6: you don't want a world where just a few companies 250 00:13:21,240 --> 00:13:22,480 Speaker 6: are controlling. 251 00:13:22,000 --> 00:13:23,360 Speaker 3: It, but that is the world we live in. 252 00:13:24,000 --> 00:13:25,800 Speaker 4: I don't think so, I think more and more. 253 00:13:25,840 --> 00:13:29,040 Speaker 6: What we're seeing is that with open source you can 254 00:13:29,200 --> 00:13:32,480 Speaker 6: actually distribute power more and you want to reduce the 255 00:13:32,520 --> 00:13:35,719 Speaker 6: gap between the most powerful companies and the rest of 256 00:13:35,760 --> 00:13:40,719 Speaker 6: the world, not only other companies but policy makers, non profits, 257 00:13:40,920 --> 00:13:42,439 Speaker 6: academia and all of that. 258 00:13:42,600 --> 00:13:44,080 Speaker 4: And that's the purpose of open source. 259 00:13:44,120 --> 00:13:47,680 Speaker 6: It reduces the gap between the most powerful companies and 260 00:13:47,720 --> 00:13:50,280 Speaker 6: the rest of the world, and that's what creates kind 261 00:13:50,280 --> 00:13:54,880 Speaker 6: of like a sustainable, balanced future for AI and technology. 262 00:13:55,160 --> 00:13:57,000 Speaker 5: It's a conversation where you're going to have all day 263 00:13:57,040 --> 00:14:00,520 Speaker 5: long and you point out that basically open source allows 264 00:14:01,360 --> 00:14:02,960 Speaker 5: more groups to go. 265 00:14:02,920 --> 00:14:03,600 Speaker 4: To work on it. 266 00:14:03,679 --> 00:14:07,880 Speaker 5: The problem is, as we're learning the tens of billions 267 00:14:07,880 --> 00:14:12,480 Speaker 5: of dollars it takes to train models with yes, tens 268 00:14:12,520 --> 00:14:15,440 Speaker 5: or hundreds of billions of parameters, and then you go 269 00:14:15,520 --> 00:14:18,680 Speaker 5: lower down and now what we're hearing is that actually 270 00:14:18,679 --> 00:14:20,920 Speaker 5: there are the folks doing this running out of cash? 271 00:14:21,120 --> 00:14:22,640 Speaker 5: Are you seeing that as well? 272 00:14:23,040 --> 00:14:25,560 Speaker 6: So it's more less true because for example, now you 273 00:14:25,600 --> 00:14:30,360 Speaker 6: can use Lamastri that really has been closely but that 274 00:14:30,880 --> 00:14:34,880 Speaker 6: meta has released and finds units for a very small 275 00:14:35,040 --> 00:14:37,840 Speaker 6: amount of money. That's why I'm hugging face. There's over 276 00:14:38,080 --> 00:14:41,680 Speaker 6: one million models that have been trained by companies, and 277 00:14:41,720 --> 00:14:43,840 Speaker 6: a lot of these companies are very small and don't 278 00:14:43,840 --> 00:14:46,520 Speaker 6: have like really really big big budget. I feel like 279 00:14:46,640 --> 00:14:50,880 Speaker 6: today every single company has to build their own AI 280 00:14:51,840 --> 00:14:55,800 Speaker 6: otherwise they run the risk of being left behind. And 281 00:14:55,840 --> 00:14:58,840 Speaker 6: that's what we're seeing, and it doesn't require any more 282 00:14:58,840 --> 00:15:02,440 Speaker 6: really really big budget. An interesting point though, that we're 283 00:15:02,480 --> 00:15:05,400 Speaker 6: going to see this year though, is that we'll need 284 00:15:05,440 --> 00:15:10,760 Speaker 6: to find for AI companies better business models. That's what 285 00:15:11,120 --> 00:15:14,280 Speaker 6: kind of like you hinted at, something we really focused 286 00:15:14,280 --> 00:15:17,240 Speaker 6: on at Talking Face. We are looking grateful to be 287 00:15:17,520 --> 00:15:22,000 Speaker 6: close to profitable, which is very unusual for AI companies. 288 00:15:22,520 --> 00:15:25,760 Speaker 6: But we're starting to see that there are some ways 289 00:15:25,800 --> 00:15:32,680 Speaker 6: to generate revenue and not burn insane amounts of compute 290 00:15:33,520 --> 00:15:34,720 Speaker 6: for AI startups today. 291 00:15:34,800 --> 00:15:38,720 Speaker 3: I mean, how on that profitability perspective of yours, how 292 00:15:38,760 --> 00:15:40,600 Speaker 3: many paying customers do you now have Can you give 293 00:15:40,640 --> 00:15:42,320 Speaker 3: us an update. You've got a million models. What about 294 00:15:42,320 --> 00:15:43,040 Speaker 3: paying customers? 295 00:15:43,280 --> 00:15:47,360 Speaker 6: We have more than ten thousand paying customers out of 296 00:15:47,440 --> 00:15:51,120 Speaker 6: the over one hundred thousand organizations, more than four million 297 00:15:51,200 --> 00:15:54,480 Speaker 6: AI builders that are using our platform, and I think 298 00:15:54,520 --> 00:15:59,760 Speaker 6: we found the right balance between monetizing, especially with like 299 00:16:00,160 --> 00:16:03,240 Speaker 6: big companies that are using the platform in. 300 00:16:03,200 --> 00:16:05,520 Speaker 4: Private enterprise companies exactly in. 301 00:16:05,560 --> 00:16:10,040 Speaker 6: Order to fund all the free community, open source work 302 00:16:10,400 --> 00:16:12,760 Speaker 6: that we're doing, and that is always going to stay 303 00:16:12,840 --> 00:16:13,760 Speaker 6: open source and free. 304 00:16:13,760 --> 00:16:15,040 Speaker 3: Of course, I want to go back to what I 305 00:16:15,160 --> 00:16:17,080 Speaker 3: was saying though about people running out of cash. You 306 00:16:17,080 --> 00:16:22,080 Speaker 3: actually put out a really interesting call on x basically saying, look, 307 00:16:22,120 --> 00:16:23,920 Speaker 3: I'm here if you need me. Hugging face is here. 308 00:16:23,920 --> 00:16:26,520 Speaker 3: If there are good people out there building interesting businesses 309 00:16:26,520 --> 00:16:28,080 Speaker 3: but you're running out of money, we could be a 310 00:16:28,120 --> 00:16:32,120 Speaker 3: home for you. Are you making acquisitions? Is it acquihiring 311 00:16:32,160 --> 00:16:32,960 Speaker 3: that goes on them? 312 00:16:33,400 --> 00:16:36,239 Speaker 6: We make some acquisitions. We're going to have interesting announcements 313 00:16:36,280 --> 00:16:37,000 Speaker 6: in the next few weeks. 314 00:16:37,040 --> 00:16:39,160 Speaker 3: I oh, don't taaser, but that's interesting. 315 00:16:39,400 --> 00:16:41,920 Speaker 6: But I think in geneoin AI you're going to see 316 00:16:41,920 --> 00:16:45,120 Speaker 6: more and more MNA because, as you said, I think 317 00:16:45,160 --> 00:16:49,160 Speaker 6: a lot of companies took very risky bets. A lot 318 00:16:49,200 --> 00:16:51,640 Speaker 6: of them are running out of money. And at the 319 00:16:51,680 --> 00:16:54,200 Speaker 6: same time you have other companies like Hugging Face and 320 00:16:54,280 --> 00:16:59,080 Speaker 6: others that are successful enough to be homes. Some of 321 00:16:59,120 --> 00:17:01,040 Speaker 6: these M and A is going to weird, right, We've 322 00:17:01,040 --> 00:17:04,879 Speaker 6: seen that happening in Lidit with Deck with some. 323 00:17:05,520 --> 00:17:09,879 Speaker 5: Usual marriage with necessity rather than choice. 324 00:17:10,040 --> 00:17:10,480 Speaker 4: May's. 325 00:17:10,760 --> 00:17:10,920 Speaker 6: Yes. 326 00:17:11,680 --> 00:17:14,440 Speaker 5: One thing that's good about summits like these Bloomberg Tech 327 00:17:14,720 --> 00:17:16,080 Speaker 5: by the way, we can go around the room and 328 00:17:16,119 --> 00:17:17,840 Speaker 5: ask who you're going to be shopping for. That's going 329 00:17:17,920 --> 00:17:20,359 Speaker 5: to be interesting. But you get all these people in 330 00:17:20,359 --> 00:17:24,680 Speaker 5: one place. You've also used the time that you've been 331 00:17:24,680 --> 00:17:28,560 Speaker 5: in San Francisco because you're up in Seattle, right, Miami 332 00:17:28,640 --> 00:17:32,639 Speaker 5: or Miami Apologies, you've been hiring, you've been interviewing candidates. 333 00:17:33,440 --> 00:17:35,680 Speaker 5: Is that just a function of the best candidates being 334 00:17:35,760 --> 00:17:39,600 Speaker 5: here in this city? How wide are you casting your net? 335 00:17:39,880 --> 00:17:40,080 Speaker 4: Yeah? 336 00:17:40,119 --> 00:17:42,119 Speaker 6: I think I think San Francisco is still the heart 337 00:17:42,440 --> 00:17:45,240 Speaker 6: of technology and the I right, there's so much talent, 338 00:17:45,320 --> 00:17:49,720 Speaker 6: so much so many interesting companies, so many interesting big 339 00:17:49,760 --> 00:17:53,119 Speaker 6: technology companies being here that it's important for us to 340 00:17:53,240 --> 00:17:56,160 Speaker 6: kind of like you have a foot on the ground here. 341 00:17:56,200 --> 00:17:58,679 Speaker 6: We have a team already here, but we're also hiring 342 00:17:59,359 --> 00:18:01,600 Speaker 6: community for hugging face here. 343 00:18:02,640 --> 00:18:03,720 Speaker 3: Applications being taken. 344 00:18:04,080 --> 00:18:09,200 Speaker 6: Yes, there's a really massive fight and struggle for AI 345 00:18:09,440 --> 00:18:16,679 Speaker 6: talents right now with inflation of packages everywhere. But what 346 00:18:16,720 --> 00:18:18,800 Speaker 6: we're seeing is that when when you have a mission 347 00:18:18,960 --> 00:18:23,160 Speaker 6: that's like interesting two candidates like we open source, then 348 00:18:23,200 --> 00:18:25,399 Speaker 6: you can attract really good talents. That's one of the 349 00:18:25,440 --> 00:18:28,119 Speaker 6: reasons why. Also we're seeing big tech doing more and 350 00:18:28,119 --> 00:18:31,000 Speaker 6: more open source. Right if you look at Meta with 351 00:18:31,119 --> 00:18:32,040 Speaker 6: all the great work. 352 00:18:31,880 --> 00:18:34,960 Speaker 3: That favorites, you keep on. I mean you've only really 353 00:18:34,960 --> 00:18:37,680 Speaker 3: mentioned number three. You're trying to cajole Google into coming 354 00:18:37,720 --> 00:18:38,720 Speaker 3: even more open source. 355 00:18:39,200 --> 00:18:42,640 Speaker 6: I think as long as companies contribute to the world 356 00:18:42,680 --> 00:18:45,160 Speaker 6: and to the field, if we've open source, we open research, 357 00:18:45,920 --> 00:18:48,680 Speaker 6: I think it benefits everyone. I think we've we've lost 358 00:18:48,680 --> 00:18:50,920 Speaker 6: a little bit this way in the US for the 359 00:18:50,960 --> 00:18:54,200 Speaker 6: past few years. If you look at AI five years ago, 360 00:18:54,359 --> 00:18:57,080 Speaker 6: most of it was open source and open science. It 361 00:18:57,200 --> 00:18:59,360 Speaker 6: changed a little bit when some companies started to make 362 00:18:59,440 --> 00:19:03,119 Speaker 6: money and and changing their approach to things. But I 363 00:19:03,119 --> 00:19:05,040 Speaker 6: think it would be positive for the world to get 364 00:19:05,080 --> 00:19:08,440 Speaker 6: back to an AI domain that is more open, more transparent, 365 00:19:08,560 --> 00:19:09,280 Speaker 6: more inclusive. 366 00:19:09,560 --> 00:19:11,560 Speaker 5: And you asked a calendar date now, by the way, 367 00:19:11,680 --> 00:19:13,600 Speaker 5: because you told us you're going to announce. 368 00:19:13,280 --> 00:19:16,440 Speaker 4: The news when you're ready four weeks. Yeah, we're holding you. 369 00:19:16,880 --> 00:19:19,760 Speaker 3: Thanks clems on, our joy to have you with us 370 00:19:19,800 --> 00:19:22,280 Speaker 3: and let you get to your breakfast where he's holding 371 00:19:22,320 --> 00:19:36,080 Speaker 3: court here seeo hugging face. What a great conversation. Welcome 372 00:19:36,080 --> 00:19:38,600 Speaker 3: back to this very special edition of Blue Meg Technology, 373 00:19:38,680 --> 00:19:41,720 Speaker 3: Live in the Heart in San Francisco. All the grain 374 00:19:41,800 --> 00:19:44,160 Speaker 3: and the good of industry movers shakers when it comes 375 00:19:44,160 --> 00:19:47,520 Speaker 3: to artificial intelligence, in particular the academics, but the companies 376 00:19:47,560 --> 00:19:51,520 Speaker 3: behind it, the CEOs, and notably also the investors. And 377 00:19:51,560 --> 00:19:53,399 Speaker 3: this is an interesting one for the investor base. Right, 378 00:19:53,480 --> 00:19:56,159 Speaker 3: we potentially have a new large language model getting at 379 00:19:56,160 --> 00:19:56,960 Speaker 3: a decent evaluation. 380 00:19:57,080 --> 00:19:57,280 Speaker 4: Yeah. 381 00:19:57,320 --> 00:19:59,280 Speaker 5: So, I think what we reported last night is that 382 00:19:59,640 --> 00:20:03,639 Speaker 5: xa I, the AI company started by Elon Musk and 383 00:20:04,359 --> 00:20:07,400 Speaker 5: which he built out pretty quickly, is closing this kind 384 00:20:07,400 --> 00:20:12,400 Speaker 5: of mega funding round eighteen billion dollar valuation. The thing 385 00:20:12,480 --> 00:20:15,000 Speaker 5: is that we've learned right over the last year or 386 00:20:15,080 --> 00:20:18,480 Speaker 5: more that is not actually that eyewatering. A number the 387 00:20:18,560 --> 00:20:20,679 Speaker 5: numbers involved are not that iwak it wasn't he. 388 00:20:21,000 --> 00:20:24,280 Speaker 3: Actually raising an awful lot considering an eighteen billion valuation. 389 00:20:24,440 --> 00:20:27,400 Speaker 5: Yes, I think we've reported sort of up to six 390 00:20:27,440 --> 00:20:30,399 Speaker 5: billion dollars. The thing is that the compute costs a 391 00:20:30,480 --> 00:20:33,399 Speaker 5: mega and I took a phone call this morning saying, 392 00:20:33,920 --> 00:20:36,520 Speaker 5: look past the cash and start asking whether the XAI 393 00:20:36,520 --> 00:20:39,000 Speaker 5: has got access to the GPUs. Now Elon Musk has 394 00:20:39,040 --> 00:20:43,160 Speaker 5: an existing relationship with Nvidio Jensen hung in the Tesla context, 395 00:20:43,160 --> 00:20:45,800 Speaker 5: but it's a good gossip for the Bloomberg Tec event. 396 00:20:45,840 --> 00:20:47,480 Speaker 4: It's a good thing to discuss well. 397 00:20:47,520 --> 00:20:50,119 Speaker 3: And ultimately who are the investors? What we've seen I 398 00:20:50,160 --> 00:20:52,160 Speaker 3: think really the rise in twenty twenty three and twenty 399 00:20:52,160 --> 00:20:53,879 Speaker 3: twenty four, it's been corporate VC. 400 00:20:54,600 --> 00:20:56,560 Speaker 4: Yes, of course, a strategic investor. 401 00:20:56,720 --> 00:20:59,679 Speaker 3: Yeah, you've got Sequoia Capitals being incredibly active, who were 402 00:20:59,720 --> 00:21:01,840 Speaker 3: going to have to co capital on a little bit later. 403 00:21:02,080 --> 00:21:03,480 Speaker 3: But then more at the seed of the funding the 404 00:21:03,480 --> 00:21:07,000 Speaker 3: series A series people. Amount of money necessary for these 405 00:21:07,080 --> 00:21:09,200 Speaker 3: large language les means and video has to be a 406 00:21:09,240 --> 00:21:10,440 Speaker 3: player or a Google has. 407 00:21:10,600 --> 00:21:12,760 Speaker 5: And what I heard is that Jared Birchall, whose head 408 00:21:12,760 --> 00:21:14,680 Speaker 5: of Musk's family office, has been in the Middle East 409 00:21:14,720 --> 00:21:16,240 Speaker 5: tolling the sovereigns a lot of that. 410 00:21:16,240 --> 00:21:16,959 Speaker 4: Stuff going on. 411 00:21:27,400 --> 00:21:30,800 Speaker 3: Welcome back to a special edition of Bloomberg Technology right 412 00:21:30,840 --> 00:21:33,119 Speaker 3: here in San Francisco. An event is upon our hands 413 00:21:33,119 --> 00:21:35,399 Speaker 3: that has all to do with artificial intelligence and what 414 00:21:35,680 --> 00:21:38,280 Speaker 3: to continue that conversation? Right here, right now at the 415 00:21:38,320 --> 00:21:42,600 Speaker 3: Bloomberg Tech Summit is writer CEO Mayhabim, who joins us now, 416 00:21:42,640 --> 00:21:46,399 Speaker 3: who has been doing Janata AI for the enterprise before 417 00:21:46,400 --> 00:21:48,560 Speaker 3: everyone else got with the program. You both rate in 418 00:21:48,600 --> 00:21:51,080 Speaker 3: twenty twenty, you've got an enormous chunk of change from 419 00:21:51,520 --> 00:21:54,679 Speaker 3: Iconic Capital and other key investors. And how does it 420 00:21:54,720 --> 00:21:57,080 Speaker 3: feel with everyone trying to surge in on the enterprise 421 00:21:57,160 --> 00:22:01,400 Speaker 3: opportunity here? How are you standing out? I'm making sure 422 00:22:01,400 --> 00:22:04,199 Speaker 3: that you keep the keys like ubers and clients that 423 00:22:04,240 --> 00:22:04,720 Speaker 3: you already have. 424 00:22:05,359 --> 00:22:08,560 Speaker 7: Yeah, it's actually really exciting to see all of the investment. 425 00:22:08,680 --> 00:22:10,719 Speaker 7: Right We've been working on this, my Covaner and I 426 00:22:10,800 --> 00:22:13,840 Speaker 7: for ten years previously in a machine translation startup, and 427 00:22:13,920 --> 00:22:17,200 Speaker 7: so to see all of this attention is actually amazing. 428 00:22:17,280 --> 00:22:19,960 Speaker 7: But the way we stand out, I think, is with 429 00:22:20,080 --> 00:22:24,600 Speaker 7: a really differentiated platform that helps enterprises with the last mile, 430 00:22:24,680 --> 00:22:26,560 Speaker 7: which is ninety percent of the work in AI. 431 00:22:27,560 --> 00:22:29,520 Speaker 5: May You've been on the show a number of times 432 00:22:29,520 --> 00:22:32,840 Speaker 5: over the last two years or so, and each time 433 00:22:32,880 --> 00:22:35,800 Speaker 5: I always reflect on sort of the rate of change 434 00:22:35,840 --> 00:22:39,160 Speaker 5: for the industry, but also grow for your company. Clem 435 00:22:39,160 --> 00:22:41,520 Speaker 5: DeLong of Hugging Face just gave us some numbers about 436 00:22:41,560 --> 00:22:44,160 Speaker 5: the sort of size and scope of how they're doing 437 00:22:44,440 --> 00:22:47,280 Speaker 5: if you say close to profit or near profit or 438 00:22:47,320 --> 00:22:49,840 Speaker 5: something like that, but just tell us about your company 439 00:22:49,840 --> 00:22:50,480 Speaker 5: and how it's doing. 440 00:22:50,760 --> 00:22:53,600 Speaker 7: Yeah, I mean, it's been an incredible rate of change. 441 00:22:54,040 --> 00:22:56,639 Speaker 7: When we started the company, we knew AI was going 442 00:22:56,680 --> 00:22:59,359 Speaker 7: to be better at people at reading and writing, and 443 00:22:59,400 --> 00:23:02,520 Speaker 7: that has certainly happened. We now say, if you can 444 00:23:02,560 --> 00:23:04,960 Speaker 7: write it, you can build it, because AI is not 445 00:23:05,000 --> 00:23:08,080 Speaker 7: just the technology, it's the way to build new technology. 446 00:23:08,440 --> 00:23:12,479 Speaker 7: But building AI apps is actually still quite difficult, and 447 00:23:12,520 --> 00:23:15,080 Speaker 7: so the rate of change of just what we've been 448 00:23:15,119 --> 00:23:17,960 Speaker 7: able to do, I mean, it's hundreds of enterprise customers, 449 00:23:18,080 --> 00:23:22,280 Speaker 7: hundreds of thousands of users, thousands of applications that are 450 00:23:22,320 --> 00:23:26,120 Speaker 7: in production. So a lot of this kind of question around, 451 00:23:26,320 --> 00:23:30,639 Speaker 7: like how you get applications from POC to scale. You know, 452 00:23:30,680 --> 00:23:32,639 Speaker 7: we've been doing that for years now and it's just 453 00:23:32,720 --> 00:23:34,560 Speaker 7: had a tremendous impact on the growth. 454 00:23:34,359 --> 00:23:34,879 Speaker 3: Of the business. 455 00:23:35,200 --> 00:23:39,119 Speaker 5: You have some relatively new work on models, right, so 456 00:23:39,280 --> 00:23:42,160 Speaker 5: tell us about the kind of the latest and greatest 457 00:23:42,200 --> 00:23:43,520 Speaker 5: on the tech side of your offerings. 458 00:23:43,640 --> 00:23:43,880 Speaker 8: Yeah. 459 00:23:44,240 --> 00:23:47,560 Speaker 7: So, over the past few months, we've introduced vision as 460 00:23:47,960 --> 00:23:52,200 Speaker 7: a capability into a platform. We've launched Palmyra in thirty 461 00:23:52,240 --> 00:23:56,399 Speaker 7: two languages that really really high quality, beating human benchmarks 462 00:23:56,400 --> 00:23:59,119 Speaker 7: our customers tell us. And next up for us our 463 00:23:59,200 --> 00:24:03,440 Speaker 7: large reasoning models, so software that write software, which we're 464 00:24:03,480 --> 00:24:06,479 Speaker 7: really excited about being able to go from you know, 465 00:24:06,560 --> 00:24:11,560 Speaker 7: work substitution to real work reinvention and orchestration using AI. 466 00:24:12,359 --> 00:24:15,639 Speaker 3: At the very start, you said basically ninety percent of 467 00:24:15,640 --> 00:24:18,280 Speaker 3: the work isn't just getting the right language, large language 468 00:24:18,320 --> 00:24:20,520 Speaker 3: model in the door, but it's implementing it. It's all 469 00:24:20,560 --> 00:24:22,480 Speaker 3: the other bells and whistles that go to ensure that 470 00:24:22,520 --> 00:24:24,879 Speaker 3: you get operational efficiencies that you put it to your 471 00:24:24,920 --> 00:24:28,360 Speaker 3: own workflows. What are some of the best ways you're 472 00:24:28,359 --> 00:24:30,760 Speaker 3: seeing and being harnessed. What are some of the worst ways, 473 00:24:30,800 --> 00:24:33,760 Speaker 3: Because everyone's still waiting for this Eureka moment where all 474 00:24:33,760 --> 00:24:36,399 Speaker 3: of our exuberants around AI actually makes a real difference. 475 00:24:36,440 --> 00:24:39,800 Speaker 7: Around one hundred percent, there are fifteen hundred lms, right, 476 00:24:39,800 --> 00:24:41,840 Speaker 7: if large languine hundred Yeah, I mean, and. 477 00:24:41,800 --> 00:24:43,680 Speaker 3: They can pass the MCAT and the LSAT. 478 00:24:43,680 --> 00:24:46,320 Speaker 7: So if lllms were the answer, everyone would have the 479 00:24:46,359 --> 00:24:48,919 Speaker 7: generative AI program of their dreams, right, But that's not 480 00:24:49,040 --> 00:24:52,000 Speaker 7: the case. There's so much work to get the data 481 00:24:52,080 --> 00:24:55,240 Speaker 7: and the context and the workflow from the business user 482 00:24:55,359 --> 00:24:58,800 Speaker 7: into the application, right, And that's what our platform does. 483 00:24:58,840 --> 00:25:02,320 Speaker 7: It's this collaborative inner that combines the LLM with all 484 00:25:02,359 --> 00:25:05,360 Speaker 7: of those building blocks, and that's where the magic is 485 00:25:05,480 --> 00:25:08,960 Speaker 7: because the llms themselves need so much more context about 486 00:25:08,960 --> 00:25:11,800 Speaker 7: the business to be able to do what customers need 487 00:25:11,840 --> 00:25:12,159 Speaker 7: them to know. 488 00:25:12,840 --> 00:25:15,439 Speaker 3: You said before that basically large language models are going 489 00:25:15,480 --> 00:25:18,399 Speaker 3: to be commoditized. The foundational models are going to be commoditized, 490 00:25:18,400 --> 00:25:22,080 Speaker 3: particularly from a consumer perspective. Where then does the value 491 00:25:22,160 --> 00:25:25,320 Speaker 3: ultimately end up lying? Because there are so many people 492 00:25:25,359 --> 00:25:27,639 Speaker 3: trying to fix problems using generative AI, A lot of 493 00:25:27,640 --> 00:25:30,240 Speaker 3: them are coming to you to try and be bought 494 00:25:30,280 --> 00:25:32,320 Speaker 3: or helped at the moment, I assume because they're running 495 00:25:32,320 --> 00:25:35,960 Speaker 3: out of money themselves. Yeah, there's certainly a lot of air. 496 00:25:35,880 --> 00:25:38,120 Speaker 7: Being sucked out of a room by big tech, right, 497 00:25:38,240 --> 00:25:41,520 Speaker 7: but there's still a ton of opportunity for startups. Microsoft 498 00:25:41,680 --> 00:25:44,919 Speaker 7: has to build for the lowest common denominator, right, so 499 00:25:45,240 --> 00:25:49,600 Speaker 7: individual productivity is very different than team productivity and team workflows. 500 00:25:49,840 --> 00:25:53,400 Speaker 7: So even though it feels like we're going to go 501 00:25:53,440 --> 00:25:57,760 Speaker 7: through sort of a big consolidation phase, I do think 502 00:25:57,760 --> 00:26:00,480 Speaker 7: there's still a ton of opportunity for for stars. We 503 00:26:00,560 --> 00:26:03,520 Speaker 7: have made a small acquisition that will announce soon, and 504 00:26:03,560 --> 00:26:06,679 Speaker 7: I think we'll make others. So there certainly is a 505 00:26:06,760 --> 00:26:10,640 Speaker 7: real high barrier for entry to come in and serve 506 00:26:10,680 --> 00:26:13,879 Speaker 7: the enterprise. But it's still there's so much blank white 507 00:26:13,920 --> 00:26:17,040 Speaker 7: open space for startups to help enterprises compete. 508 00:26:17,200 --> 00:26:20,240 Speaker 5: It's interesting maybe that you use the c word consolidation. 509 00:26:20,320 --> 00:26:22,560 Speaker 5: I don't think glem DeLong went as far as using 510 00:26:22,560 --> 00:26:26,400 Speaker 5: the word consolidation, but I think you know, you said 511 00:26:26,440 --> 00:26:28,960 Speaker 5: something a moment ago about big tech sucking the oxygen 512 00:26:29,000 --> 00:26:31,880 Speaker 5: out of the room. It goes to the open source 513 00:26:32,000 --> 00:26:37,080 Speaker 5: closed debate. I assume you sit on the open source side, 514 00:26:37,400 --> 00:26:38,560 Speaker 5: but just wigh in. 515 00:26:38,800 --> 00:26:40,000 Speaker 3: So we're kind of in the middle. 516 00:26:40,119 --> 00:26:43,639 Speaker 7: So our models are proprietary, A bunch are on hugging 517 00:26:43,640 --> 00:26:47,800 Speaker 7: face so later generations of models, but our latest models 518 00:26:47,800 --> 00:26:48,800 Speaker 7: are are closed source. 519 00:26:48,880 --> 00:26:50,520 Speaker 3: But by being in the middle. 520 00:26:50,680 --> 00:26:55,400 Speaker 7: What enterprises really need is the ability to audit right 521 00:26:55,480 --> 00:26:59,040 Speaker 7: and have the transparency around training data and all sorts 522 00:26:59,080 --> 00:27:02,359 Speaker 7: of things related to the models they don't really want, 523 00:27:02,560 --> 00:27:07,520 Speaker 7: like the last mile cumbersomeness of necessarily like fine tuning 524 00:27:07,560 --> 00:27:10,560 Speaker 7: or running the models themselves is what we're finding. And 525 00:27:10,600 --> 00:27:13,760 Speaker 7: so like in the in the kind of sucking the 526 00:27:13,800 --> 00:27:17,640 Speaker 7: air out of the room, the confusion around what vendors 527 00:27:17,680 --> 00:27:20,760 Speaker 7: to turn to and how to actually get great applications shift. 528 00:27:21,040 --> 00:27:22,960 Speaker 7: That's where That's where I think there's still a lot 529 00:27:23,000 --> 00:27:25,880 Speaker 7: of confusion in the enterprise, and I think there's still 530 00:27:25,880 --> 00:27:29,320 Speaker 7: all that work to be done to minimize hallucinations. 531 00:27:28,440 --> 00:27:32,199 Speaker 3: To ensure that we're seeing a clarity of where the 532 00:27:32,320 --> 00:27:35,520 Speaker 3: underlying data is coming from and you're not having copyright issues. 533 00:27:35,880 --> 00:27:38,880 Speaker 3: Give us clarity on your business. Now, have you been 534 00:27:38,920 --> 00:27:43,399 Speaker 3: approached to be bought? Are you remaining independent? Are you 535 00:27:43,520 --> 00:27:44,320 Speaker 3: raising more money? 536 00:27:44,880 --> 00:27:50,840 Speaker 7: So there's there's a really long, i think, product journey 537 00:27:50,880 --> 00:27:53,720 Speaker 7: for us to really realize our vision. So I'm really 538 00:27:53,760 --> 00:27:56,720 Speaker 7: excited about remaining independent. It used to be a year 539 00:27:56,760 --> 00:27:59,680 Speaker 7: ago that I would say, you know, lms are for 540 00:27:59,800 --> 00:28:01,600 Speaker 7: the rudgery, the work you don't want. 541 00:28:01,400 --> 00:28:02,119 Speaker 3: To do today. 542 00:28:02,440 --> 00:28:04,960 Speaker 7: The capabilities are so incredible, they're as good as us. 543 00:28:05,400 --> 00:28:09,000 Speaker 7: But the future is work where you get to do 544 00:28:09,280 --> 00:28:11,560 Speaker 7: the work you want to do and lllms do the rest, 545 00:28:11,640 --> 00:28:14,560 Speaker 7: right because one person's drudgery is another person's creative passion, 546 00:28:14,800 --> 00:28:16,600 Speaker 7: and that's kind of compelling. 547 00:28:16,200 --> 00:28:17,679 Speaker 3: Vision for the future of work. 548 00:28:17,880 --> 00:28:21,240 Speaker 7: We're not seeing enterprises come up with. Yes, we talk 549 00:28:21,280 --> 00:28:23,600 Speaker 7: to hundreds of companies a week, and that really feels 550 00:28:23,640 --> 00:28:26,840 Speaker 7: missing right now. Kind of executives painting a vision for 551 00:28:26,920 --> 00:28:29,160 Speaker 7: what AI looks like inside their companies in a way 552 00:28:29,200 --> 00:28:32,199 Speaker 7: that brings people along. So there's a lot to do 553 00:28:32,400 --> 00:28:34,920 Speaker 7: both in you know, kind of bringing our vision into 554 00:28:34,960 --> 00:28:37,080 Speaker 7: the world and helping companies achieve theirs. 555 00:28:37,920 --> 00:28:38,080 Speaker 8: Right. 556 00:28:38,160 --> 00:28:40,720 Speaker 5: A CEO may have be great to catch up here 557 00:28:40,960 --> 00:28:43,000 Speaker 5: at Blue veg Tech in San Francisco. 558 00:28:43,320 --> 00:28:46,360 Speaker 3: She slies every week from some Franciscot and on and 559 00:28:46,360 --> 00:28:46,720 Speaker 3: I'm back. 560 00:28:47,040 --> 00:28:49,480 Speaker 5: That's what we're hearing, the world of the CEO in 561 00:28:49,520 --> 00:28:52,160 Speaker 5: the world of AI on a plane coming out here. 562 00:28:52,200 --> 00:28:54,240 Speaker 4: We're going to be joined by Stephanie Jang. 563 00:28:54,320 --> 00:28:59,080 Speaker 5: Partner at Sequoia, for her take on investing in AI startups. 564 00:28:59,120 --> 00:29:13,800 Speaker 5: Stay with us, we'll be right back. This is Bloomberg Technology. 565 00:29:16,240 --> 00:29:19,720 Speaker 5: Welcome back to this special edition of Bloomberg Technology. We're 566 00:29:19,760 --> 00:29:23,480 Speaker 5: back together live in San Francisco a Bloomberg Tech, our 567 00:29:23,520 --> 00:29:26,600 Speaker 5: annual conference, and here at the Tech Summit, we've got 568 00:29:26,640 --> 00:29:31,120 Speaker 5: to talk about investing the first checks into those new 569 00:29:31,240 --> 00:29:34,240 Speaker 5: and early AI startups. We have a fantastic guest for 570 00:29:34,280 --> 00:29:39,400 Speaker 5: today's Visa Spotlight, Stephanie Jan, partner at Sequoia. You guys 571 00:29:39,440 --> 00:29:43,440 Speaker 5: are so busy, you are writing lots of checks, but 572 00:29:45,000 --> 00:29:48,280 Speaker 5: the new companies being founded in AI are not the 573 00:29:48,360 --> 00:29:51,200 Speaker 5: same as they were one year ago, and certainly not 574 00:29:51,280 --> 00:29:52,240 Speaker 5: eighteen months or. 575 00:29:52,240 --> 00:29:52,880 Speaker 4: Two years ago. 576 00:29:53,320 --> 00:29:55,160 Speaker 5: Just give us the sort of timeline of where we 577 00:29:55,240 --> 00:29:58,080 Speaker 5: are now in this industry way. 578 00:29:59,440 --> 00:30:01,040 Speaker 1: First of all, thank you so much for having me 579 00:30:01,120 --> 00:30:04,320 Speaker 1: at Caroline. It's an absolute joy to be here. We're 580 00:30:04,360 --> 00:30:07,640 Speaker 1: at a really interesting time in AI today, seven years 581 00:30:07,640 --> 00:30:10,920 Speaker 1: from the advent of the transformer, four years. 582 00:30:10,640 --> 00:30:13,080 Speaker 3: Since the advent of the GPT three moment. 583 00:30:13,600 --> 00:30:15,440 Speaker 1: I think twenty twenty four is going to be a 584 00:30:15,480 --> 00:30:17,080 Speaker 1: monumental year for AI. 585 00:30:17,480 --> 00:30:19,960 Speaker 3: And here's why. I think this year is. 586 00:30:19,920 --> 00:30:23,040 Speaker 1: Going to be a step function leap in digital intelligence, 587 00:30:23,360 --> 00:30:27,440 Speaker 1: everything from video to AI agents to robotics. I also 588 00:30:27,480 --> 00:30:29,160 Speaker 1: think that this year is going to be the year 589 00:30:29,240 --> 00:30:32,400 Speaker 1: we see a shift in the ecosystem to a thriving 590 00:30:32,440 --> 00:30:36,960 Speaker 1: ecosystem with many winners in the models area across closed source, 591 00:30:37,080 --> 00:30:40,640 Speaker 1: open source, large models, small models, and third, I also 592 00:30:40,680 --> 00:30:42,400 Speaker 1: think this is the year we start to see AI 593 00:30:42,440 --> 00:30:44,000 Speaker 1: commercialization at scale. 594 00:30:44,280 --> 00:30:46,040 Speaker 3: And at Sequoia, we've been really busy. 595 00:30:46,080 --> 00:30:49,000 Speaker 1: As you noted, we're highly selective about the companies that 596 00:30:49,040 --> 00:30:51,680 Speaker 1: we partner with, but this year, in just the first 597 00:30:51,760 --> 00:30:55,680 Speaker 1: four months alone, we've invested in ten new AI companies, 598 00:30:56,000 --> 00:30:59,720 Speaker 1: everything from new foundation models to new AI native applications. 599 00:31:00,040 --> 00:31:01,960 Speaker 3: I love being went through the history like seven years 600 00:31:02,000 --> 00:31:04,480 Speaker 3: ago since the transmission model. I mean it was twenty 601 00:31:04,520 --> 00:31:07,320 Speaker 3: years ago just over that Sequoia wrote the first check 602 00:31:07,760 --> 00:31:09,960 Speaker 3: into in video, and now we think there's still that 603 00:31:10,040 --> 00:31:13,320 Speaker 3: company really owning really the oxygen in the room. 604 00:31:13,240 --> 00:31:16,560 Speaker 5: And the value vest right they writing checks of their 605 00:31:16,600 --> 00:31:18,560 Speaker 5: own is a strategic investment. 606 00:31:18,120 --> 00:31:21,280 Speaker 3: And I'm interested therefore exactly to AT's point, how competitive 607 00:31:21,320 --> 00:31:23,960 Speaker 3: is it out there to get those first checks in 608 00:31:24,240 --> 00:31:26,920 Speaker 3: ho Who are you seeing coming? Is it the corporates 609 00:31:26,920 --> 00:31:28,920 Speaker 3: that are wanting to write checks? Is it VC's wanted 610 00:31:28,960 --> 00:31:29,760 Speaker 3: to write checks. 611 00:31:30,240 --> 00:31:34,320 Speaker 1: It's an incredible ecosystem right now, with everyone pouring money 612 00:31:34,320 --> 00:31:37,680 Speaker 1: into the AI ecosystem. I think it's very much reflective 613 00:31:37,800 --> 00:31:41,160 Speaker 1: of the opportunity that we see in AI, the large 614 00:31:41,200 --> 00:31:44,320 Speaker 1: market opportunity that is to come. I actually think that 615 00:31:44,400 --> 00:31:46,680 Speaker 1: we're still in the very early innings. 616 00:31:46,320 --> 00:31:46,920 Speaker 3: Of all of this. 617 00:31:47,360 --> 00:31:50,760 Speaker 1: Well, you know, it's the classic saying of we overestimate 618 00:31:50,800 --> 00:31:53,480 Speaker 1: in the short run, but we really underestimate in. 619 00:31:53,480 --> 00:31:56,320 Speaker 3: The long run. And video has done a wonderful. 620 00:31:56,000 --> 00:31:59,160 Speaker 1: Job of being such a critical hold in the ecosystem 621 00:31:59,520 --> 00:32:02,280 Speaker 1: with hard we're driving compute, but also now with so 622 00:32:02,320 --> 00:32:05,880 Speaker 1: many software tools and the entire developer ecosystem they've built 623 00:32:05,920 --> 00:32:08,320 Speaker 1: around them. So I think that we're just in the 624 00:32:08,360 --> 00:32:10,040 Speaker 1: early innings and there's a lot more to come. 625 00:32:10,560 --> 00:32:14,720 Speaker 3: We were just speaking with Clem from Hugging Face Anadeine Mayhembib, 626 00:32:14,760 --> 00:32:17,880 Speaker 3: who highlight the fact that it's really expensive to do 627 00:32:17,960 --> 00:32:21,719 Speaker 3: this and video chips are a putty penny. How are 628 00:32:21,720 --> 00:32:25,400 Speaker 3: you seeing the companies that you back able to sustain 629 00:32:25,520 --> 00:32:27,160 Speaker 3: the investment they need to make. How do you make 630 00:32:27,200 --> 00:32:28,760 Speaker 3: sure the checks you write you're going in the right 631 00:32:28,800 --> 00:32:30,800 Speaker 3: direction and not just sort of going into the pool 632 00:32:30,840 --> 00:32:32,240 Speaker 3: of training money. 633 00:32:32,560 --> 00:32:36,160 Speaker 1: Yeah, well, I think that the classic conventional wisdom is 634 00:32:36,200 --> 00:32:40,520 Speaker 1: that incumbents with scale, data, capital and distribution have a 635 00:32:40,600 --> 00:32:44,560 Speaker 1: natural advantage, and that's absolutely correct. It also costs a 636 00:32:44,560 --> 00:32:46,840 Speaker 1: lot to build these models because of compute and for 637 00:32:46,920 --> 00:32:49,400 Speaker 1: AI talent, but I also think that there are so 638 00:32:49,520 --> 00:32:53,520 Speaker 1: many nimble ways for a startup to compete. Specifically, I 639 00:32:53,560 --> 00:32:57,520 Speaker 1: think the next leap is really around one high quality data, 640 00:32:57,640 --> 00:33:03,440 Speaker 1: specifically high quality labels of data and targeted domain specific data. 641 00:33:03,640 --> 00:33:06,080 Speaker 3: Second, it's really about what you do with that data. 642 00:33:06,800 --> 00:33:10,080 Speaker 1: Reinforcement learning with human feedback I think will really shine 643 00:33:10,080 --> 00:33:10,880 Speaker 1: in this next era. 644 00:33:11,360 --> 00:33:12,200 Speaker 3: It's an idea. 645 00:33:12,040 --> 00:33:15,960 Speaker 1: Derived from reinforcement learning, but here an agent actually also 646 00:33:16,120 --> 00:33:19,160 Speaker 1: learns on the fly with human feedback, and that's what's 647 00:33:19,280 --> 00:33:23,320 Speaker 1: so brilliant about chat topt for example. And finally, I 648 00:33:23,360 --> 00:33:26,280 Speaker 1: think that you really differentiate not just on model performance, 649 00:33:26,360 --> 00:33:29,240 Speaker 1: which is where all the capital goes into, but it's 650 00:33:29,280 --> 00:33:32,880 Speaker 1: also around product distribution and the entire product experience that 651 00:33:32,920 --> 00:33:33,720 Speaker 1: you offer. 652 00:33:33,840 --> 00:33:34,840 Speaker 3: To the end customer. 653 00:33:35,800 --> 00:33:38,120 Speaker 5: You use the word incumbent, I think we should probably 654 00:33:38,120 --> 00:33:41,760 Speaker 5: talk about who those incumbents are because the point that 655 00:33:42,160 --> 00:33:44,920 Speaker 5: may have even right made to a certain extent claim 656 00:33:44,920 --> 00:33:47,360 Speaker 5: from Hogeyface is that big tech and where I think 657 00:33:47,360 --> 00:33:51,360 Speaker 5: we're talking about alphabet Microsoft in the first instance, are 658 00:33:51,400 --> 00:33:54,680 Speaker 5: sucking the oxygen out of the room. From a capital perspective, 659 00:33:54,680 --> 00:33:58,560 Speaker 5: a talent perspective, you invest in the preceed and seed sage. 660 00:33:59,120 --> 00:34:01,800 Speaker 4: Do you find that be true? 661 00:34:01,840 --> 00:34:04,840 Speaker 1: Well, I think that incumbents absolutely have an advantage, as 662 00:34:04,840 --> 00:34:07,480 Speaker 1: we just outlined, but I also think that new startups 663 00:34:07,520 --> 00:34:11,360 Speaker 1: have a shop scales. At AI actually recently released the 664 00:34:11,480 --> 00:34:14,560 Speaker 1: survey last week where they interviewed thousands of developers on 665 00:34:14,680 --> 00:34:17,879 Speaker 1: their most popular models, and the ones that actually came 666 00:34:18,000 --> 00:34:21,640 Speaker 1: into light were GPT four, GPT three point five, and 667 00:34:21,719 --> 00:34:25,040 Speaker 1: Gemini as the most popular models used. But we're also 668 00:34:25,080 --> 00:34:28,320 Speaker 1: starting to see new players come into play with models 669 00:34:28,320 --> 00:34:31,960 Speaker 1: that are just as competitive in performance. I'm really excited 670 00:34:31,960 --> 00:34:35,839 Speaker 1: about the open source model ecosystem enabling many more new 671 00:34:35,840 --> 00:34:39,040 Speaker 1: players to come into play. LAMA three, for instance, is 672 00:34:39,080 --> 00:34:42,600 Speaker 1: so powerful. The new eight billion PARAMETERAR model is a 673 00:34:42,880 --> 00:34:46,440 Speaker 1: longer trained, small model that I think will become a 674 00:34:46,480 --> 00:34:50,000 Speaker 1: really powerful building block for new developers to build new 675 00:34:50,040 --> 00:34:53,480 Speaker 1: applications on top of and to build new models around it. 676 00:34:53,480 --> 00:34:56,400 Speaker 1: It's going to drastically reduce the cost of what it 677 00:34:56,440 --> 00:34:57,959 Speaker 1: takes to build new experiences. 678 00:34:58,280 --> 00:35:03,360 Speaker 5: We are increasingly talking about Beta and its competence in 679 00:35:03,840 --> 00:35:08,319 Speaker 5: building large language models. You speak highly of them. Where 680 00:35:08,360 --> 00:35:11,120 Speaker 5: do you see them they? I think, Zuckerberg said on 681 00:35:11,160 --> 00:35:14,760 Speaker 5: the Cool last week, we want to be the world's 682 00:35:14,880 --> 00:35:16,080 Speaker 5: leading AI company. 683 00:35:16,920 --> 00:35:18,400 Speaker 4: Where are they in that journey? 684 00:35:19,400 --> 00:35:22,759 Speaker 1: I think that they have an incredible advantage, and not 685 00:35:22,920 --> 00:35:25,239 Speaker 1: just because of the capital that they're willing to pour 686 00:35:25,320 --> 00:35:28,399 Speaker 1: into play, but also because of the entire treasure trove 687 00:35:28,480 --> 00:35:32,280 Speaker 1: of data that they hold, all this proprietary UGC content 688 00:35:32,320 --> 00:35:35,640 Speaker 1: that they can really use to train their models. One 689 00:35:35,640 --> 00:35:38,080 Speaker 1: of the things I'm really excited to see them enter 690 00:35:38,120 --> 00:35:40,960 Speaker 1: the scene with this year is a new generative video 691 00:35:41,000 --> 00:35:42,799 Speaker 1: foundation model, similar. 692 00:35:42,560 --> 00:35:44,680 Speaker 3: To what we saw with Sora and open Ai. 693 00:35:45,440 --> 00:35:48,719 Speaker 1: To me, the most powerful thing that unlocked was that 694 00:35:48,800 --> 00:35:52,279 Speaker 1: the methodology we take for building large language models and 695 00:35:52,360 --> 00:35:56,279 Speaker 1: digital intelligence works for video as well. You take a 696 00:35:56,280 --> 00:35:59,439 Speaker 1: diffusion transformer model and you just scale it with enough 697 00:35:59,520 --> 00:36:03,480 Speaker 1: video dat and compute and meta has a wonderful advantage 698 00:36:03,600 --> 00:36:06,239 Speaker 1: given the entire treasure trow of content they have. 699 00:36:06,680 --> 00:36:07,960 Speaker 3: To compete in the bosom. 700 00:36:08,760 --> 00:36:10,680 Speaker 1: And then what they're doing with Lama three I think 701 00:36:10,800 --> 00:36:14,840 Speaker 1: is game changing entirely. It opens the playing field for 702 00:36:15,000 --> 00:36:19,720 Speaker 1: everyone themselves new startups, lowering the cost for a thriving 703 00:36:19,760 --> 00:36:21,160 Speaker 1: ecosystem with many winners. 704 00:36:21,640 --> 00:36:24,080 Speaker 3: Come back when you've got more checks you can announce 705 00:36:24,200 --> 00:36:26,560 Speaker 3: in that thriving ecosystem. Such a joy to be here 706 00:36:26,560 --> 00:36:28,480 Speaker 3: with you. Thank you so much for having it, Caroline, 707 00:36:28,480 --> 00:36:41,920 Speaker 3: and having by Sequoia partner Stephanie Chan. Welcome back to 708 00:36:41,960 --> 00:36:45,680 Speaker 3: this special edition of Bloomberg Technology, the heart of San Francisco, 709 00:36:46,080 --> 00:36:48,799 Speaker 3: big event upon our hands and every year in fact, 710 00:36:49,000 --> 00:36:52,160 Speaker 3: Rumbag Business Week releases in tandem. It's a list of 711 00:36:52,239 --> 00:36:55,160 Speaker 3: tech wants to watch. But these are the startup founders, 712 00:36:55,239 --> 00:36:59,520 Speaker 3: the big tech managers, the mom he investors as well, 713 00:36:59,680 --> 00:37:02,400 Speaker 3: who of playing a big role in shaping text future 714 00:37:02,760 --> 00:37:05,319 Speaker 3: and joining us now is one of these ones. To 715 00:37:05,360 --> 00:37:08,720 Speaker 3: words please to welcome you did madame Amazon vice president 716 00:37:09,120 --> 00:37:11,880 Speaker 3: for of course the worldwide operations side of the business, 717 00:37:12,320 --> 00:37:16,200 Speaker 3: your first interviews. It's taking on an enormous role of 718 00:37:16,239 --> 00:37:20,200 Speaker 3: more than a million people that you manage the focus 719 00:37:20,239 --> 00:37:23,080 Speaker 3: of getting my package to me in the swiftest way, 720 00:37:23,160 --> 00:37:26,239 Speaker 3: most cost efficient manner as possible. Can I just ask 721 00:37:26,280 --> 00:37:29,239 Speaker 3: what your day looks like? What is a day in like? 722 00:37:29,880 --> 00:37:32,640 Speaker 8: Well, first, Caroline and thank you for having me it's 723 00:37:32,640 --> 00:37:33,480 Speaker 8: supposed to be here. 724 00:37:33,920 --> 00:37:34,120 Speaker 5: Yeah. 725 00:37:34,239 --> 00:37:37,680 Speaker 8: For me, really, my day starts, you know, fairly early 726 00:37:37,719 --> 00:37:39,800 Speaker 8: in the morning. But you know, it starts with thinking about, 727 00:37:40,040 --> 00:37:41,680 Speaker 8: you know, the team I've got. 728 00:37:41,760 --> 00:37:44,319 Speaker 4: You know, we've got a very very broad team all 729 00:37:44,360 --> 00:37:45,080 Speaker 4: around the world. 730 00:37:45,600 --> 00:37:48,520 Speaker 8: We've got four thousand different locations that we operate around 731 00:37:48,520 --> 00:37:50,960 Speaker 8: the world, and really it's focused on how do we 732 00:37:51,000 --> 00:37:54,480 Speaker 8: continue to innovate on behalf of customers that do it 733 00:37:54,520 --> 00:37:56,440 Speaker 8: in a way that puts safety. 734 00:37:56,400 --> 00:37:57,680 Speaker 4: And people at the forefront. 735 00:37:58,080 --> 00:38:01,920 Speaker 8: And so my day is really focused on innovation across 736 00:38:02,120 --> 00:38:03,359 Speaker 8: four different spectrums. 737 00:38:03,600 --> 00:38:05,319 Speaker 4: Safety, really the. 738 00:38:05,320 --> 00:38:10,360 Speaker 8: Customer experience with delivery speeds innovating, especially with what's happening 739 00:38:10,360 --> 00:38:13,440 Speaker 8: with technology finding new ways, you know, whether that's through 740 00:38:13,480 --> 00:38:16,560 Speaker 8: robotics our operations to make things more efficient and driving 741 00:38:16,560 --> 00:38:17,480 Speaker 8: you when you're. 742 00:38:17,320 --> 00:38:20,120 Speaker 5: Talking about the technology, we're talking about everything from the fleets, 743 00:38:20,200 --> 00:38:23,520 Speaker 5: right so there's a transition to sustainable energy in the 744 00:38:23,600 --> 00:38:28,120 Speaker 5: fleet context, talking about robotics in the fulfillment centers, and 745 00:38:28,239 --> 00:38:31,520 Speaker 5: dare I say AI in tracking the data? What's the 746 00:38:31,560 --> 00:38:35,200 Speaker 5: biggest investment focused for you right now? And technology roll out? 747 00:38:35,560 --> 00:38:38,520 Speaker 8: You know, we've got technology all across our operations and 748 00:38:38,560 --> 00:38:41,680 Speaker 8: there's really two things that I would maybe thematically talk about. 749 00:38:41,800 --> 00:38:43,960 Speaker 8: One is, we do have a lot of investments in 750 00:38:44,000 --> 00:38:46,920 Speaker 8: automation and robotics that are going on, especially with how 751 00:38:47,000 --> 00:38:50,360 Speaker 8: quickly things are accelerating with General VII. We have investments 752 00:38:50,520 --> 00:38:55,000 Speaker 8: on really novel foundational models that look and use the 753 00:38:55,080 --> 00:38:57,600 Speaker 8: high quality data that we've gathered in source as we 754 00:38:57,640 --> 00:38:59,879 Speaker 8: ship tens of millions of products every day, and those 755 00:39:00,040 --> 00:39:01,920 Speaker 8: going to help make some of those robotic solutions more 756 00:39:01,920 --> 00:39:04,480 Speaker 8: generalizable as well as make them more efficient. 757 00:39:04,840 --> 00:39:05,800 Speaker 4: And the second is we've. 758 00:39:05,680 --> 00:39:09,719 Speaker 8: Been working on a set of really inventive robotic solutions 759 00:39:09,719 --> 00:39:12,120 Speaker 8: over the last few years that are finally reaching maturity 760 00:39:12,160 --> 00:39:14,680 Speaker 8: and scale and we'll start to roll out starting this year. 761 00:39:14,960 --> 00:39:17,120 Speaker 8: Both those are really exciting and it will be transformative 762 00:39:17,160 --> 00:39:17,520 Speaker 8: for operation. 763 00:39:17,560 --> 00:39:19,959 Speaker 3: I mean, you've got to be inventive because Annie Jase 764 00:39:20,120 --> 00:39:23,640 Speaker 3: is asking you to focus on costs, but I'm sure 765 00:39:23,640 --> 00:39:26,000 Speaker 3: the innovation in a way does longer term once you 766 00:39:26,080 --> 00:39:28,400 Speaker 3: made the investment strip out some of the costs, but 767 00:39:28,480 --> 00:39:30,680 Speaker 3: ultimately does that come at a sacrifice of labor. How 768 00:39:30,760 --> 00:39:32,400 Speaker 3: do you talk to those people that you are so 769 00:39:32,560 --> 00:39:34,719 Speaker 3: key when you focused on to ensure that they feel 770 00:39:34,719 --> 00:39:36,360 Speaker 3: that are being augmented not replaced. 771 00:39:36,680 --> 00:39:38,920 Speaker 8: You know, the best thing I can talk about is 772 00:39:38,960 --> 00:39:42,160 Speaker 8: our history. You've deployed seven hundred and fifty thousand robots 773 00:39:42,160 --> 00:39:45,400 Speaker 8: over the last decade across our operations. We've done that 774 00:39:45,440 --> 00:39:48,239 Speaker 8: while creating hundreds of thousands of jobs. And you know 775 00:39:48,280 --> 00:39:50,400 Speaker 8: what's really interesting and not a lot of people know, 776 00:39:50,520 --> 00:39:54,320 Speaker 8: is we've created dozens of new classic jobs, your skilled jobs, 777 00:39:54,440 --> 00:39:57,600 Speaker 8: technical jobs. And what we've learned in that process is 778 00:39:57,640 --> 00:39:59,799 Speaker 8: that one of the most important things that you can 779 00:39:59,880 --> 00:40:03,120 Speaker 8: do you as a company in this world of generatively 780 00:40:03,239 --> 00:40:07,040 Speaker 8: I and robotics, is to really focus on investing employees. 781 00:40:07,040 --> 00:40:10,279 Speaker 8: So we've launched two different programs. One is a twenty 782 00:40:10,320 --> 00:40:14,120 Speaker 8: twenty five off skilling pledge that really helps train people 783 00:40:14,480 --> 00:40:17,839 Speaker 8: for this new workplace in the future, and a Yeah 784 00:40:17,960 --> 00:40:21,080 Speaker 8: Ready program that's generally available to everybody that's really focused 785 00:40:21,120 --> 00:40:24,440 Speaker 8: on investing in helping provide a skill training. 786 00:40:24,239 --> 00:40:25,280 Speaker 4: Over two million people. 787 00:40:25,480 --> 00:40:29,040 Speaker 8: So really focusing on people alongside the investments who are 788 00:40:29,040 --> 00:40:29,520 Speaker 8: making in general. 789 00:40:29,600 --> 00:40:31,360 Speaker 4: VI in Revidy, we just have thirty seconds. 790 00:40:31,400 --> 00:40:33,840 Speaker 5: What's your one personal goal for the year, something you 791 00:40:33,840 --> 00:40:34,400 Speaker 5: want to achieve. 792 00:40:34,719 --> 00:40:37,160 Speaker 8: You know, for me, there's more than one, but I'll 793 00:40:37,239 --> 00:40:39,360 Speaker 8: quickly I'll try to answer it quickly. The first and 794 00:40:39,400 --> 00:40:41,120 Speaker 8: the highest priority for us is safety, and we want 795 00:40:41,160 --> 00:40:43,520 Speaker 8: to be the safest workplace across the industries we operate 796 00:40:43,560 --> 00:40:47,719 Speaker 8: in making measurable and really remarkable progress in that area. 797 00:40:47,719 --> 00:40:49,239 Speaker 8: I want to company to invest in that. And the 798 00:40:49,280 --> 00:40:52,560 Speaker 8: second is to compete to improve the convenience for customers 799 00:40:53,000 --> 00:40:54,720 Speaker 8: and delivery speeds is an area of focus. 800 00:40:55,880 --> 00:40:59,120 Speaker 3: Congratulations on being one of the key ones to watch. 801 00:40:59,480 --> 00:41:02,080 Speaker 3: Phenomenal the amount of people who manage young age that 802 00:41:02,080 --> 00:41:04,759 Speaker 3: you are, madame. We thank you, Amazon vice President of 803 00:41:04,880 --> 00:41:09,200 Speaker 3: Worldwide Operations. Meanwhile, I mean from ones to watch of 804 00:41:09,200 --> 00:41:11,760 Speaker 3: individuals to everything you've got to watch coming up, because 805 00:41:11,800 --> 00:41:14,160 Speaker 3: this is going to be an amazing set of conversations. 806 00:41:14,440 --> 00:41:16,960 Speaker 3: I'm going to be speaking with a key chip leader. 807 00:41:17,160 --> 00:41:19,360 Speaker 3: Of course, you're going to be speaking about the future 808 00:41:19,360 --> 00:41:20,799 Speaker 3: and technology. Who have you got lined up? 809 00:41:20,840 --> 00:41:22,440 Speaker 5: Yeah, I'm going to talk to Tom Oxley of synchron 810 00:41:22,480 --> 00:41:24,759 Speaker 5: I'm going to talk about brain implants and what the 811 00:41:24,840 --> 00:41:28,239 Speaker 5: right method of putting a electrode into one's brain is. 812 00:41:29,160 --> 00:41:33,040 Speaker 3: I love asual casuals perspective. Renee James is joining me 813 00:41:33,080 --> 00:41:35,520 Speaker 3: and Perco. Look, this is the question that having just 814 00:41:35,560 --> 00:41:38,440 Speaker 3: spoken with Renee the other Renee and chips of arm. 815 00:41:38,480 --> 00:41:42,080 Speaker 3: Where is the market share being taken by these newer players, 816 00:41:42,160 --> 00:41:45,600 Speaker 3: taking from AMD, from Intel, even potentially in video. 817 00:41:46,239 --> 00:41:48,280 Speaker 5: Thank you so much for joining us on this special 818 00:41:48,400 --> 00:41:51,440 Speaker 5: edition of Bloomberg Technology. It's great to be back together 819 00:41:51,480 --> 00:41:53,920 Speaker 5: in the field, but we actually have a full day ahead, 820 00:41:54,000 --> 00:41:56,440 Speaker 5: so many great guests stay with us. Thank you for 821 00:41:56,480 --> 00:42:00,719 Speaker 5: tuning in from San Francisco at Bloomberg Tech for this 822 00:42:00,880 --> 00:42:01,480 Speaker 5: is Bloomberg