Latest / Newcomer Pod / Inside Cerebral Valley: Autonomous Vehicles & AI Investment
Transcript
- 0:00This is brought to you by forethought building AI agents
- 0:03for every customer moment. Hey, it's Madeline from the
- 0:07newcomer podcast. We're fresh off the Serbo Valley
- 0:10AI Summit in London, and we're going to bring you two of our
- 0:12liveliest on stage discussions into the podcast feed.
- 0:16Both touch on one of the biggest themes of the event, which was
- 0:18where AI will enter our lives beyond the chatbot, whether it's
- 0:21through voice on a device or even in our cars.
- 0:25First up, we'll hear from Uber CEO Dara Khosrowshahi and Alex
- 0:29Kendall from Wave, who are teaming up to bring self driving
- 0:31cars to the UK. They actually rode in a Wave
- 0:34self driving car to the summit, which was really cool to see.
- 0:37The next panel is a conversation between some of the top VCs in
- 0:40Europe, Philippe Bateri from Excel, Jan Hammer from Index
- 0:44Ventures, and Tom Holm from Google Ventures about where
- 0:47they're seeing the next big opportunities for AI founders
- 0:49globally. Give it a listen.
- 0:56Hey, well, we're down to the last two, two of the discussions
- 1:00I'm most excited about. So excited to be here with Dara
- 1:04and Alex. I, you know, covered all the
- 1:08Uber periods. I remember riding in an Uber
- 1:11self driving car in 2016, pre Dara when felt like there was a
- 1:16lot of hype, a lot of interventions and then Uber had
- 1:23to pull them off the road because I think they maybe went
- 1:25a little too early. And now, you know, almost a
- 1:28decade later, I think I've gone from being a self driving bear
- 1:33to being really excited about the opportunity.
- 1:36So I just want to start off. I mean your two companies are
- 1:39are working together to bring self driving cars to London.
- 1:44Dara, you want to start off why? Why work with Wave and what is
- 1:47the partnership going to look like here?
- 1:48Well, we're working with Wave because they're an absolute
- 1:51leader in the field. Our strategy at at Uber is we
- 1:56are no longer developing our own self self driving technology.
- 1:59We want to really work with the ecosystem and look for the best
- 2:03and brightest. We're developing self driving
- 2:04technology. We've built a platform that
- 2:08brings incredible demand on a global basis.
- 2:11We can help these highly technical companies with
- 2:14operations, local operations, etcetera and we can be a vehicle
- 2:20to bring the self driving technology and introduce it to
- 2:23consumers. And so we have a rare view into
- 2:30all of the development that is happening in the industry and
- 2:35Wave is unquestionably one of the leaders.
- 2:37I think they're taking A and Alice can talk to it kind of a
- 2:41bit of a differentiated approach in terms of an end to end large
- 2:47model. And once we saw what they were
- 2:50doing, once we experienced the the driving efficacy of of their
- 2:56driver and how it can generalize.
- 2:58Ashley, Alice and I were driven here by a wave driver from Wave
- 3:03headquarters. With a safety.
- 3:05Driver right with a safety driver.
- 3:07Any interventions or 0. How was the drive?
- 3:10Perfect, perfect. And, and there were some really,
- 3:12really complex maneuvers that we had to go through construction
- 3:17trucks parked on the left, so it had to go around trucks.
- 3:23Pedestrians in London being pedestrians in London.
- 3:26It was fairly. I love all all the look left,
- 3:29look right on the ground. You really need the
- 3:31instructions. But, but these guys have been a
- 3:33real leader and so we, we very much wanted to work with them.
- 3:36We've invested in the company and and we're very much looking
- 3:38forward to bringing their technology to London and then
- 3:43and then beyond. Maybe you can talk a little bit
- 3:44about that. And.
- 3:45I mean, regulators have said, oh, you could do it next year.
- 3:48I'm sure you guys were supportive of that.
- 3:49But are you prepared to do it next year?
- 3:52How do you feel? How ready are you to deliver
- 3:56self driving cars with Uber in London?
- 3:58Well, look, I think the thing that's very exciting for us is
- 4:01the confidence and clarity we have from regulators here.
- 4:04So we're working as hard as we can to get this launched in a
- 4:07way that scales of course in London, but really worldwide.
- 4:11And I, I can't wait to do that on the, on the Uber network.
- 4:14But the, the clarity and conviction that the UK
- 4:17government has said, look, we want to make this possible.
- 4:19We want to accelerate regulation by a bit over a year and make
- 4:23this possible from next year. It's, it's, it's great for us
- 4:25because we can say, OK, we can commit to building this out here
- 4:28and it won't, you know, it won't be a full service all of a
- 4:32sudden from day zero. There will be of course steps
- 4:34that we take in a controlled and responsible way.
- 4:37But ultimately. You know what neighborhood will
- 4:38be first. Well, I, we've all got our
- 4:41favourites, but I that, that, that debate's still ongoing.
- 4:45But I think being able to what we enable with end to end AI is
- 4:49of course to move away from the concept of high definition maps.
- 4:52You don't need to map each region.
- 4:54Actually, the AI has the kind of intelligence to understand
- 4:57scenarios it's never seen before.
- 4:58My favorite experience from our drive, we went through this
- 5:00intersection where there was a, a car that stopped in front of
- 5:03us and we had to navigate. I wait for an oncoming car and I
- 5:06lift safely, change lanes to get around them and keep the traffic
- 5:09flowing. And all of this kind of
- 5:11complexity our AI can handle naturally.
- 5:13And I think it's not just about having a a, you know, AI driver
- 5:18that gets you from A to B, but keeps the traffic flowing in a
- 5:20way that doesn't cause any road angst around you.
- 5:22I think that's what's going to be really loved by London, as a
- 5:25new experience that fits in with a London driving culture as well
- 5:28as being of course. Are you using Lidar or what's
- 5:31your philosophy in terms of sensors?
- 5:33We build an AI foundation model that can address all ranges of
- 5:36autonomy. So the car we drove in today was
- 5:39camera only. Although the what?
- 5:41Kind of car was it? It was a Ford.
- 5:43Marquee. But the vehicles we target for
- 5:45the Level 4 robotaxi applications, they're going to
- 5:49be have redundant sensors. So camera, radar lighter.
- 5:51The key thing though is that they're not going to have a
- 5:54retrofitted spinning lighter on the roof, but integrated sensors
- 5:57that the manufacturers can produce at mass volume.
- 6:00And I think that's going to make this economically viable and
- 6:02scalable. I think just one comment on
- 6:05what's really cool about the UK and the regulators actually
- 6:07accelerating the timeline here is there are very few times when
- 6:11you have a true platform revolution the way that you're
- 6:15seeing with AI. And I do think that from a
- 6:18European standpoint, there could be the view that the US has been
- 6:24the country that has benefited from, you know, the big platform
- 6:27changes, Internet, broadband, mobile, cloud, etcetera.
- 6:32It has been the US companies that have grabbed a significant
- 6:36share of the enormous opportunity that has been
- 6:39created as a result of these platform changes.
- 6:42AII mean you've heard that this is something that that you're
- 6:45deeply in represents such a platform change and the talent
- 6:51level that we see in the UK coming out of the educational
- 6:55institutions is incredible. And so if the UK has an
- 7:00opportunity to be a leader here in terms of this change and kind
- 7:03of the application of AI generally, physical AI.
- 7:06Are we at a point yet where cities they should almost be
- 7:09bidding for the opportunity or like, you know, in San
- 7:11Francisco, it feels like tourists, tourists are trying to
- 7:14go see way MO's. Like, do you do you get a sense
- 7:17that the mood in cities has pivoted it all to, oh man, being
- 7:22first is like a compelling pitch?
- 7:24We're we're seeing cities leaned in now you should be leaned in
- 7:28at the same time, safety is paramount.
- 7:32The cost of a mistake in the real world is so much higher
- 7:36than a hallucination in the digital world.
- 7:39So while they are leaned in, these are long dialogues that
- 7:44we're having with regulators to make sure that they you can be
- 7:48leaned in but safe at the same time.
- 7:51But so far the conversations that we're that we've had are
- 7:53very, very constructive. But we have to meter how we how
- 7:58we lean in here. I mean, what we see is that
- 7:59city, this solves so many problems for cities, but it's of
- 8:02course it comes down to the nuance of how you integrate and
- 8:05deploy this technology. And of course, doing it in a
- 8:07responsible way is. Critical.
- 8:09How do you prove it to London? Or what's the path to showing
- 8:11that you are safe? There's so many aspects.
- 8:15To. Regulation.
- 8:17And the key thing for us is of course we prove it before we
- 8:19deploy. And so of course that starts
- 8:22with very large scale offline testing and simulation, right?
- 8:25It starts with proving that you can expose the AI to the right
- 8:29amount of data across the right behaviors and then testing the
- 8:32integration. I believe the Texas approach is
- 8:35prove it after you deploy, we'll get more on that.
- 8:38Dara, I mean you are partnering with other self driving car
- 8:42providers, you're working with way MO like what what do you
- 8:46think about sort of your partnership strategy and how do
- 8:49you decide which partners to work with?
- 8:51Well, ultimately we think this is a technology that holds the
- 8:55promise of making the streets of the world safer.
- 8:58That is by far the most important factor here in the US
- 9:01alone, there are probably 35,000 automotive fatalities per year.
- 9:06And if you take the opportunity that we have now, which is to
- 9:09hold the standard of safety of these AI drivers being multiple
- 9:13times better than humans, and Waymo certainly has demonstrated
- 9:17that that's possible. You get to have safer streets
- 9:20and over a longer period of time, you can bring the cost of
- 9:23transportation down, increase the man and hopefully take a
- 9:26chunk out of personal car ownership there.
- 9:30So that's the that's what ultimately the opportunity to
- 9:34represents. We don't think there's going to
- 9:36be a winner take all in the marketplace.
- 9:38And so we want to partner with the best and brightest.
- 9:43We get to meet everybody in in the industry.
- 9:47We have a safety case as well that we want to make sure is
- 9:51fully satisfied. And Wave is just unquestionably
- 9:55one of our top partners that we've met and we're just super,
- 9:59super excited about building the relationship.
- 10:00I mean you're in sort of a funny position in the like a lot of
- 10:04people in this room and probably Alex all want the technology of
- 10:07AI to really be the differentiator in sort of
- 10:10pricing and you sort of want to be the demand aggregation and
- 10:14sort of customer relationship is a differentiator.
- 10:17Do you agree with that or what do you see as sort of your
- 10:22leverage in in sort of deploying self driving?
- 10:26I mean, listen, our, our leverage if you want to call
- 10:27that is that we're operating in over 70 countries.
- 10:30We and so we can bring instant demand to any technology
- 10:34provider. And I think at least for the
- 10:37next 5 to 10 years, there's going to be more demand for
- 10:41these robot cars, if you want to call it that than supply.
- 10:46So 10 years from now, maybe doubly competition, who's going
- 10:49to win etcetera, and who's going to be the provider for various
- 10:51OEMs. But we are a long way away from that.
- 10:55And right now it's about making sure that we introduce this
- 10:58technology safely in the streets of, you know, hopefully all the
- 11:03countries in which we operate. So I just don't think we're
- 11:05there yet at this point. Yeah, far away, Alex, the, you
- 11:10know, there's a third mystery party in this partnership, which
- 11:13is the OEM. I don't know that you're going
- 11:16to give it up right here on stage, but like what
- 11:18characteristics do you look for? Do you think it'll be just one?
- 11:22And yeah, how do you see the roles of the three different
- 11:25players? Yeah, That's a good question,
- 11:27Eric. I mean, we work with automakers
- 11:29around the world through Europe, Japan and the United States.
- 11:33And I, I again, I think this is going to be a technology that
- 11:37all OEMs will want to produce vehicles that have autonomy
- 11:40capabilities. There will still be some
- 11:43personal car ownership market and of course the ability to
- 11:46drive hands off and eyes off with that is going to be a
- 11:48really key thing for that product experience and that's
- 11:52what our product can enable. But then the market is going to
- 11:54move to focus on level 4 robo taxis and other autonomous
- 11:57mobility applications. And So what we are seeing is the
- 12:00very best automotive manufacturers are setting
- 12:02themselves up for their future. They're working with our AI, of
- 12:05course, building vehicles that are software defined that enable
- 12:08you to the amazing thing now is that we're seeing manufacturers
- 12:11bring out vehicles that you can get data off them.
- 12:13You can we can start to aggregate data just like Uber
- 12:16aggregates demand. We aggregate data around the
- 12:18industry and that enables us to build a, you know, the most safe
- 12:22and scaled AI across the market and in the vehicles, making sure
- 12:26that there are features you need to operate a robotaxi platform.
- 12:29One OEM per city or you think there will be multiple?
- 12:33I think that consumers will want choice and we want to enable our
- 12:36platform to enable you think. The cars can provide
- 12:39differentiation too. Absolutely.
- 12:41And I mean, if you look at our AI is flexible and that it can
- 12:44work with different sensors, so it learns what a sensor
- 12:47architecture can and can't see. But some OEMs may want to put a
- 12:49camera here or slightly in a slightly different position,
- 12:52maybe they want to use a different compute stack or of
- 12:54course, the actual driving experience, the inputs to our AI
- 12:58is not just the sensor input, the navigation prompt, but also
- 13:01a style prompt. How do you want are?
- 13:04You going to give these? Are you going to be the
- 13:05scorekeeper of AI self driving? Like how much do you think
- 13:08you'll? Well, we can measure all the
- 13:10accidents. This one is slightly safer.
- 13:13We, we do have a good viewpoint as to the efficacy of the
- 13:17various AIS and the different approaches and there are
- 13:19different approaches to solving the problem here.
- 13:23And, and obviously we're going to make our bet on the
- 13:25approaches that we think are safe.
- 13:28And then also, obviously, like you said, there's a third party
- 13:31in here, which is the OEMs. I do think 10 years from now, I think
- 13:36every car sold is going to come with an L3 or L4 package
- 13:41depending on what you're looking for.
- 13:44I think it represents a huge profit opportunity for OEMs. I
- 13:48mean, like the OEMs are understanding the beauty of
- 13:50software 0 gross margin, obviously huge upfront cost in
- 13:53terms of development. So you know, every OEM that we
- 13:58talked to is very, very interested in this technology,
- 14:02but at the same time it's very different tech that they that
- 14:05they typically build. There's a big debate, do we
- 14:08build it internally, do we go with a third party, etcetera.
- 14:11But I think 10 years from now, every single new car sold, if it
- 14:14doesn't have this package available, it's not going to
- 14:16sell. And, and you think it'll you'll
- 14:19be using sort of consumer cars as part of a Uber fleet?
- 14:23Well, I think that. Or you think your cars will
- 14:25drivers. Drivers, you know, they'll drive
- 14:28our most just like popular features or cars is a Prius,
- 14:33right? And so there'll be a driver
- 14:35driving a Prius, whatever the next generation is, there'll be
- 14:38a software driver, hopefully it will be a wave driver driving
- 14:42that make of vehicle. What any thoughts on the Tesla
- 14:48roll out in Austin and what it sort of, yeah, I mean you you
- 14:53need to worry about I guess the overall reputation of self
- 14:56driving and how much do you think sort of any one player
- 15:00acting more or less recklessly will sort of blow back on the
- 15:03rest of self driving? Yeah.
- 15:05I mean, listen, I think that Tesla has been an unbelievable
- 15:08innovator in EVs. They've been innovator and and
- 15:11developing self driving software as well.
- 15:15And and certainly I wouldn't call their start reckless at
- 15:17all. I mean it's a it's a very small
- 15:19operational domain. They've got safe drivers there.
- 15:24Yeah. So I think they're, they're
- 15:25being actually appropriately careful and, and you see the
- 15:29challenges in, in real life driving there.
- 15:32So at this point, again, like this is, it's a new industry.
- 15:38We think that it's great to have so much excitement around it.
- 15:41I think when people experience it, it kind of blows them away.
- 15:44Then they start taking it for granted for like 2 minutes.
- 15:46There's absolute delight. And then they're like, you know,
- 15:49texting you on their phone. I mean, I trust it, Yeah.
- 15:51I mean I've been mostly in San Francisco Waymo's but I I trust
- 15:55them more than the average Uber drive.
- 15:57Yeah, yeah. I'm curious as to what you
- 15:58think. I mean, you, you, you're in it.
- 16:00So I'm curious. Do you trust them more than
- 16:02humans yet? I, I think, I think that's going
- 16:06to be the future is that of course they are going to be in
- 16:09so many ways, they're going to have more intelligence and be
- 16:12safer than what what we can do when we drive cars.
- 16:14But the, the operation today that we saw deployed in, in, in
- 16:17Austin from, from Tesla. I, I think the key thing from
- 16:20here is that this is going to grow.
- 16:22It's a start and there's going to be iteration from here.
- 16:25The critical thing is that as an industry, we need to make sure
- 16:27that we iterate responsibly and have the right checks and
- 16:30balances in until we prove that the level of safety can be
- 16:33better than a safe and competent human driver.
- 16:37Like talking about technology, like a core problem that I think
- 16:40we've seen with AI is that like getting very close, getting to
- 16:44like 95% feels awesome, but then it doesn't.
- 16:49Like if it doesn't solve the case, then you can't get rid of
- 16:52the humans and then it's not as valuable because they're human
- 16:54drivers. I mean, well, one of the
- 16:56interesting things there is to look at the difference between
- 16:58cognitive and embodied AI and large language models.
- 17:01Actually, how do you define safety is quite nebulous and
- 17:04there's a lot of debate around what general purpose AI
- 17:06regulation and standards should look like, but that's not true
- 17:09in, in self driving. We have very clear industry
- 17:12understanding of what safe is, of what expectations are.
- 17:14And actually the way we evaluate and benchmark these systems is
- 17:17very well understood. So that I, I, I think there is a
- 17:20push in the, in AI, we're seeing more of a push towards safety
- 17:24and benchmarks in embodied AI. And I think we're seeing
- 17:26innovation about how do we understand and prove the levels
- 17:29of these systems come out through robotics that may
- 17:33advance us as a field here. But I feel good about this in
- 17:36self driving and I I still think that conversation is evolving in
- 17:39the in the language model space. But you know, I, I'm interested
- 17:42in this beyond self driving, almost the lessons from self
- 17:45driving to some of the pieces of AI where it feels like we're
- 17:48getting answers, but then we're also getting hallucinations.
- 17:51I mean in the Uber case, I mean you came in, you did spin out
- 17:55Aurora, like how did you? We we merge ATG into Aurora.
- 17:59Yeah. And then then became its own
- 18:00company. How did you decide, I guess that
- 18:03self driving wasn't there in 2017 or they this shouldn't be
- 18:07something to invest in or how did you go into sort of all your
- 18:10technologists saying, oh, we're just tomorrow, just tomorrow and
- 18:14say I'm not convinced. Listen, I I think Eric was a
- 18:18really tough decision at the time.
- 18:19And one is we were deeply marked by the fact that we had an
- 18:23accident, lost life there. And so the responsibility that
- 18:29you hold in your hands when you're operating in the real
- 18:33world hit us deeply. I think the other circumstance
- 18:38that affected the ultimate judgement was that we, we ran
- 18:43ATG very separately from mainline Uber because we wanted
- 18:47the opportunity to partner with the rest of the ecosystem.
- 18:49Again, like we, we don't just work with, you know, we don't
- 18:53just have a Prius is on our platform.
- 18:57We got 4th. We want to work with the entire
- 18:59ecosystem. And when I went and talked with
- 19:02some of the other AV developers, I'd say, you know, ATG, it's in
- 19:06the family, but it's separate. We're going to treat you fairly.
- 19:08We're not going to share data, etc.
- 19:10You know, when you're having a conversation actually, like I
- 19:12usually do with you and I'm saying stuff and you don't
- 19:14believe a fucking can I say, pardon me?
- 19:17That's that was the response. Like the we were having
- 19:20conversations, but I think people didn't trust us.
- 19:23Yeah, because they also viewed us, rightly so, as a competitor,
- 19:26right? So we had a decision to make.
- 19:30Do we go with a proprietary kind of vertical strategy or do we
- 19:35truly partner with the industry? We couldn't have the best of
- 19:39both worlds. And ultimately we've started
- 19:41that. A partial model was the right
- 19:43model going forward and based on, you know, basically all of
- 19:47the leading players sans Tesla, but everyone else is open
- 19:52working with us, etcetera. I think it was the right
- 19:53decision to make at the time. But this was not a simple thing.
- 19:57It wasn't on a PowerPoint. It wasn't clear which way to go.
- 20:01But it was more it was more about the ability to partner
- 20:03with other people than an assessment of the technology.
- 20:06I think it was for for us it wasn't necessarily an assessment
- 20:09of technology, but it was an assessment as to what we are
- 20:12great at. And Uber is a very strong
- 20:16software company, but we're not necessarily we weren't great at
- 20:19developing self driving. We were good, but I think it
- 20:23takes better than good to get there.
- 20:26Alex, I mean, this is not monogamous partnership.
- 20:30I mean he's talking about partnering with a lot of
- 20:31different people. How do you see the partnership
- 20:33strategy and how do you get comfort partnering with a
- 20:36company that's like, oh, we want to partner with everybody?
- 20:40Well, in a similar way, the the, the platform that we're building
- 20:42is going to enable fleets and automotive manufacturers around
- 20:45the world to build autonomous products.
- 20:47But what we're seeing from Uber is leadership and speed that
- 20:49really outpaces other other fleets and marketplaces.
- 20:54Actually, it's not just the demand.
- 20:56You think about all the aspects of running on autonomy service,
- 20:59how you own, operate, clean, collect the data to integrate
- 21:01it. There's so much work there to
- 21:03take a technology works and actually scale it.
- 21:06Now like Dara, I'm a huge believer in focus and I know
- 21:11that our expertise is in building the AI model.
- 21:13We don't want to build the car. We don't want to own and operate
- 21:15it. We want to enable this AI to be
- 21:17on as many robots as possible. And and you know, that's going
- 21:20to go beyond cars in the future because I'm seeing things in
- 21:24manufacturing or other mobility spaces that are a bit like self
- 21:27driving back in 2017. And you know, it'll take some
- 21:29time before manufacturers figure out the right software to find
- 21:32vehicle equivalent. Then when they do, there's going
- 21:35to be some manufacturers like Tesla is a great example of this
- 21:38today that can build their own own vertical stack.
- 21:41But then there's going to be a wealth of, of manufacturers and
- 21:45operators that have the platform, that have the demand,
- 21:47that have the product expertise, but lack the intelligence and
- 21:50the scale that we can build there as a platform by
- 21:53aggregating data and training a model that's larger than than
- 21:57anyone can do on their own is going to unlock, I think the
- 21:59broader autonomy ecosystem. So we see it in the same way.
- 22:02But I'm, I'm so excited about, I mean, Dara's leadership and the
- 22:06speed that Hoover is pursuing here, I think we can move
- 22:09really, really quickly. Uber's worth 190 billion, which
- 22:13is great, right? I think like an all time high
- 22:16but but Tesla you know is worth a trillion you.
- 22:20Have to spoil the. Party, don't you?
- 22:21The do you think? I mean, how much do you think
- 22:23you're getting credit for your work in self driving and how
- 22:25much do you think the market should be saying Uber is a self
- 22:30driving car company? I think we're getting zero
- 22:33credit at this point. But but listen that when you're
- 22:36developing new technologies as a company, you have to be willing
- 22:39to invest aggressively for many, many years before the market
- 22:47understands or necessarily gives credit for what you're doing.
- 22:49So you know the market is bet against this before 190 billion
- 22:54isn't bad. But I think if the reality that
- 22:58we see in partnership with Wave and others and self driving
- 23:02becomes real, not just in passenger mobility but other
- 23:07kinds of mobility delivery, I think 190 will just be a step
- 23:12sum for us. Alex, you know there's a risk
- 23:16that way MO becomes synonymous with self driving cars or a lot
- 23:20of people technologists are getting the way MO experience
- 23:23first. Like where do you see your
- 23:26differentiation with them? Like how quickly do you need to
- 23:29move to sort of stay in the conversation as they keep to
- 23:32deploy and how do you assess what they have on the market
- 23:35today? Well, here's the interesting
- 23:36thing. Going city by city with a
- 23:38geofenced approach that relies on high definition maps this
- 23:41this AV. 1.0 strategy that that the industry looks at is AV. 1.0
- 23:45I like is. Is 11 play?
- 23:47But actually, I think this technology is going to see mass
- 23:51scale through personally owned vehicles first with driver
- 23:54assistance. And that's how most people are
- 23:56going to get to experience it. And I'm not clear if you can
- 23:59actually build a general purpose robotaxi without that data, that
- 24:02manufacturing integration and having our AI in millions of
- 24:06vehicles around the world for hands off and eyes off driving
- 24:09is going to provide the regulatory relationships, the
- 24:11manufacturing integration, the data, and of course, the the
- 24:14brand and the exposure around the world that we'll be able to
- 24:16take and use that to build a general purpose AI for a general
- 24:21purpose robotaxi. And so I think that is an
- 24:23enormous advantage that we can actually get that scale 1st.
- 24:27And you know, we're, we're in it for the long game.
- 24:29We want to produce something that if you think about what the
- 24:32future of robotics should be, it's not a, a robot that follows
- 24:35around infrastructure and affluent places, but we want to
- 24:38build a future where there's a level of intelligence that can
- 24:40drive in new places. It can, you know, respond to
- 24:43your demands, you can delegate tasks to it.
- 24:45It can interact with you through language.
- 24:47I mean, fundamentally, the way that's going to happen is with
- 24:49end to end deep learning. And that's the that's the bit
- 24:51we're taking. And I think that of course time
- 24:55will tell, but I think that just like we've seen in many other
- 24:57verticals from drug discovery, game playing, agents, language
- 25:01models, robotics is going to be no different.
- 25:04And I think we're going to see that better lesson play out.
- 25:07So I'm all for end to end learning being the approach that
- 25:09does scale by leveraging the scale deployment and consumer
- 25:13vehicles. Dara, do you have a year where
- 25:15you think the Uber platform flips from human majority to
- 25:19self driving majority cars? I I am 100% sure that five years
- 25:23from now we're going to need more humans on our platform than
- 25:25we do today. Business is growing really,
- 25:27really fast. 10 years from now, I'm not sure about that slope of
- 25:32the curve. It's not going to flip, but the
- 25:34increase in drivers and couriers that we need on the platform may
- 25:37start metering out and after that, who knows?
- 25:40But you're still saying 10 years away I.
- 25:42Mean I'm saying 10 years away, we will still have more drivers
- 25:47and couriers, but a higher and higher percentage of our
- 25:50platform is going to be served by by robots, one where they're
- 25:55delivery robots, drivers, etcetera.
- 25:58It's just the demand for all things on demand, whether it's a
- 26:03ride or a food delivery, grocery retail delivery, the demand is
- 26:08just growing so quickly and I think that'll continue.
- 26:11Do you do you think deployed self driving cars today are
- 26:14safer than humans? Like is is that?
- 26:16A Well, I think Waymo is demonstrating as we speak that
- 26:20it can not only be safer than a human, but it can be multiple
- 26:22times safer than a human. And I think we have an
- 26:25opportunity to hold a safety bar that is very, very high and I
- 26:30think we should take that opportunity.
- 26:32Great, Darren. Alex, thank you very much.
- 26:34Thank you. Appreciate it.
- 26:39All right, back with some of the biggest venture heavy hitters
- 26:45around, so super excited to have you guys all here.
- 26:50Yeah, obviously sort of a lot going on.
- 26:54This is really our look at the investing landscape.
- 26:58I just wanted to start off, what is a domain right now, an area,
- 27:04it probably an artificial intelligence doesn't have to be
- 27:08that you're really excited about, that you're digging into
- 27:10just like one lane that you're sort of excited about at the
- 27:14moment and you're right here. So do you want to go?
- 27:17Sure. Happy to kick off.
- 27:19I guess we've travelled a huge distance from the launch of
- 27:22ChatGPT building all the infrastructure and then the move
- 27:27to applications. I mean it's, it's been year and
- 27:29a half or so and specifically for Europe, what I'm what I'm
- 27:33excited now that we are at this application level is actually
- 27:38the traditional industries are getting changed,
- 27:40pharmaceuticals, banking and finance, lending industries like
- 27:46manufacturing, automotive, defense and so on.
- 27:49I think those industries have historically been slower to
- 27:53adopt any new technology, but now it's become really a
- 27:57necessity for them to keep up, certainly on a on a global
- 28:00scale. Applications in old school
- 28:02businesses. Are the applications replacing
- 28:05those businesses or selling to them, or both?
- 28:09AI being used as an enabler. So take a pharmaceutical
- 28:13example, which I guess Tom, you know a ton about protein
- 28:17synthesis. You could spend $100 million
- 28:20developing a new protein variant.
- 28:24You could be experimenting manually, you could be in labs,
- 28:27or you could use AI in companies like Cradle and and and and and
- 28:31others in the industry to shortcut that that cycle, that
- 28:35development cycle and reduce your cost by an order of
- 28:38magnitude. Love the answer.
- 28:41So look, the global economy is about $100 trillion and it's
- 28:44just not fully digitised yet. AI is going to unlock that low
- 28:48friction. So we've been asking like what
- 28:50applications are good enough today?
- 28:53So we started with law, invested in Harvey Law Hive coding as
- 28:58well, which these guys have phenomenal investments in as
- 29:01well. I'd give you a slightly
- 29:03different answer as well, which is I think every time you have a
- 29:06new sort of technology, there's a new interface that arises to
- 29:09sort of leverage it fully. So with the personal computer,
- 29:12you had the GUI, with the smartphone, you had touchscreen.
- 29:15And I think this is clearly going to be an era of voice,
- 29:17which is really interesting. So we've been investing in that
- 29:20space where we think voice is really high throughput.
- 29:24I look at my kids, they send voice notes on WhatsApp all the
- 29:27time because it's really quick and really efficient.
- 29:29Some of our investments in that would be really high bandwidth
- 29:32like Neuralink at the speed of thought.
- 29:35We have a stealth company in Israel that's doing sort of pre
- 29:37vocalization or silent speech which is interesting.
- 29:41And then actually investment in nothing here in London is
- 29:44related as well. I.
- 29:45Think a lot of this is voice without voice or sort of voice
- 29:48is too slow. We need to be faster than that.
- 29:51Yeah, absolutely. It's voice.
- 29:52It's, it's actually an incredibly high throughput way
- 29:56of communicating information. And so, you know, it just so
- 29:59happens that it's also probably the optimum as you've seen with
- 30:03all of the chat interfaces for LLMS for the time being.
- 30:06So we're really bullish on that space.
- 30:08Felipe. Yeah.
- 30:10I mean, I, I, I call everything that has been said here, I think
- 30:13it was very interesting with the AI to see how three years ago it
- 30:17was all about the infrastructure and the model and now it's about
- 30:21the applications. And I think in the application,
- 30:24one area which I think is very exciting right now that we
- 30:27haven't mentioned is cyber security and AI is really
- 30:31transforming cyber. Creating the problems and
- 30:33solving the problem. Exactly.
- 30:35And you have companies like Sierra who are really focusing
- 30:37on data and AI security, which really doing very well because
- 30:42that's the number one problem for CISO.
- 30:44It says, well, data is growing, data is even more valuable.
- 30:48And now it's really, it's really at risk.
- 30:51And I think after that application phase, I think
- 30:53what's going to come I think in the next 1224 months is like
- 30:57more the automation side and how do we start to automate task and
- 31:02workflow in the core enterprise. And, and right now I think we're
- 31:07seeing enterprises are dabbling with it and you have some small,
- 31:10smaller project and Pocs and I think the next 12, you know,
- 31:131224 months, I would expect much, you know, bigger chunks of
- 31:18tasks and workflow to be automated.
- 31:20I think some same way that, you know, RPA IN281718 restarted as
- 31:25a big way, we say, well, it's RPA still remain the best way to
- 31:29automate repetitive tasks, but there needs to be the same.
- 31:32But if they are the same, these are the cheapest, most efficient
- 31:36ways to do it. But then there are other tasks
- 31:38which are more complex where the outcome may not be always the
- 31:41same that that requires AI. What what is what's the mood on,
- 31:46I guess how much software is screwed right?
- 31:48You see what we had the Klarna CEO being like, oh, we're going
- 31:51to be able to build everything ourselves.
- 31:53I mean, I hear even in cyber, you know, there's almost an
- 31:56argument like we need to be so fast.
- 31:57I think I forgot if it was abnormal security I was talking
- 32:00to a couple months ago. It was like, oh, we need to move
- 32:02so quickly because people are going to be able to spin stuff
- 32:05up. I don't know just the general
- 32:07whoever wants to jump on it, but like the view on whether you can
- 32:11invest in software companies or your your own companies are
- 32:14going to build the software themselves.
- 32:15Yeah, You know, AI is here to solve a lot of these problems
- 32:19we're discussing. But back to cyber, it's also
- 32:22generating problems. Tom and I are fortunate to work
- 32:26with a company called Resistant AI and actually it's recently
- 32:30seen a massive take off because the, the surface area of fakes,
- 32:35of fraud, of falsified documents, of falsified
- 32:38identities has has expanded exponentially.
- 32:41And so at some point, you know, I recall sort of the what what I
- 32:46learned in payments a decade ago, it was it's just a cost of
- 32:49doing business. There's a certain amount of
- 32:51fraud that's always going to be there across credit card
- 32:54networks. We want to optimize that the
- 32:57total net revenue at this point, it's moved to a different
- 33:01dimension. So the problem has grown
- 33:03exponentially. And with that I think you need.
- 33:06There's some more intelligent. Solution so that doesn't.
- 33:08There's always an opportunity. Right.
- 33:10Yeah, relative thought. I think you get phases with the
- 33:14rollout of these technologies and at the moment much of their
- 33:17code is quite superficial. It's almost creating an
- 33:21explosion of technical debt. I think at the moment like you
- 33:24see amazing examples. Like they need to buy a lot of
- 33:26things are creating a mess inside their companies.
- 33:28And do you? I don't believe that software
- 33:30will be completely kind of compostable.
- 33:35It's got to stay composable. It's got to stay so that
- 33:38actually the foundations would be built.
- 33:39So the next Phase I think will be people addressing the fact
- 33:42that when you do Vibe Coder project or you do work with
- 33:46cursor and it cross references libraries or open source,
- 33:49they've got to be ones that we really believe in.
- 33:51So all three of us have invested in a company here in London
- 33:54called Tessel Guy, Pajani's new company, where he's addressing
- 33:58exactly that. Like these projects have to be
- 34:01reliable, they have to be built for the long term, and I think
- 34:03that's the next phase that will fall through.
- 34:06I think this is a real, this is a question, right?
- 34:0910 thousand $10 billion question, which is like you look
- 34:12at work day, for example, and can you say, well, it's
- 34:16tomorrow, can someone do work day?
- 34:18Or Can you imagine even a work day where basically, which is a
- 34:22very simple software, but every user will say instead of
- 34:25configuring a complex solution, we'll say, oh, I need to
- 34:28integrate with this, you know, pull the data from that, show me
- 34:31this report and dynamically the app will create the code for
- 34:35that potentially. That's the vision of the future.
- 34:39I think we're very, very long away.
- 34:40From that, I want, I don't want to spend a lot of time in this
- 34:43conference being like, why London?
- 34:44Like I, I live in New York, I host events in San Francisco.
- 34:47We're in London today. Like I think the pandemic made
- 34:50everything global and it was easy to build businesses all
- 34:52over the world. But I, I do the stats that I
- 34:55will ask the sort of why London or what the question really is.
- 34:59Like what you think the opportunity specific to London
- 35:01is and where founders should really be here.
- 35:03And then I guess the American hubris, like when should they
- 35:06really be racing to America versus focusing and local
- 35:09markets? Yeah, you would probably expect
- 35:12we would say this. It's all about talent.
- 35:14London has been the hub for not just London born and bred
- 35:18talent, but also sort of the European capital.
- 35:22Sorry to all all all all those from Amsterdam, Paris and
- 35:26Stockholm, but it is a bridge across the Atlantic and.
- 35:32How many people here are based in London?
- 35:35Raise your hand. Oh, there we go.
- 35:37And how many are not based in London?
- 35:40Minority. So there you go.
- 35:42London has a lot of talent, but also with that comes the
- 35:47ambition and and things as has been mentioned has been
- 35:51accelerated. So the speed with which
- 35:54companies are funded, speed with which companies launch and the
- 35:59speed with which they have to get to market in some in in in
- 36:03in some sense, it is that speed of iteration that is the Moat.
- 36:07Certainly in the early experimental cases where the
- 36:11customer has five competing solutions and each of those five
- 36:14providers is 1 year old, right. So you have to have speed of
- 36:17iteration, you have to raise capital fast.
- 36:19You don't have the luxury of time.
- 36:22And in some sense, that forces those companies to look big from
- 36:26the get go and, you know, get get on that bridge to you're.
- 36:30Saying if AI in particular is a race, you need to go global
- 36:33pretty fast. Tom, I want to ask you
- 36:35specifically. I mean, you're, you know, part
- 36:36of the Alphabet empire empire. Is, is DeepMind like a good
- 36:42source of startups or like has it underperformed, I guess the
- 36:46level of talent and sophistication in startup
- 36:49creation or what's what's your view on the track record in
- 36:52terms of DeepMind and spinning off startups?
- 36:56I would say the jury is still out.
- 36:57These companies are moving quickly, but the great news is
- 37:00there are world class scientists, researchers and
- 37:05applied engineers at Deep Mind that are just learning how to
- 37:09build at the very sort of Pareto front of these models and
- 37:13they're still most of them in Deep Mind and enjoying it.
- 37:16Although there's a lot of movement between all these
- 37:17companies for each of those, I think you get this kind of
- 37:20multiplier effect when they leave.
- 37:22So I don't think we've seen any of the sort of really big
- 37:25companies that we would expect come out yet.
- 37:27Maybe the one that we're most excited about is Isomorphic,
- 37:30which is the protein folding business that we span out not so
- 37:33long ago. But I went out with the leader
- 37:35of DeepMind. Or a fascinating situation.
- 37:38No, no, Demis. Well, I mean, Demis is one of
- 37:40the smartest people I've ever met.
- 37:42He works across both of those businesses and I'm confident
- 37:46that clock speed won't be kind of compromised at all.
- 37:48That's amazing. But the back to Yan's point, if
- 37:51you look at London, I don't, you know, I think building
- 37:54businesses in Europe is building them in hard mode versus the US
- 37:58But London it's less hard because you know, there's great
- 38:01capital here. The founders will start
- 38:03businesses everywhere, but for every single great founder you
- 38:06probably need 5 or 10 world class operators and there's more
- 38:09of them in London than any other city in Europe at the moment.
- 38:12Felipe, anything you'd add there, maybe either like
- 38:14categories you think in particular there's an advantage
- 38:17here or anything else and. Obviously, London has a
- 38:20privileged place in Europe, but I would say it's not the only
- 38:23ecosystem in Europe with very strong AI talents, right?
- 38:27Look at Paris. I mean, this is where found it
- 38:31fair with Yan, Lukun, etc. I mean, there's very, very great
- 38:36depth of of talent. I'm not saying this because I'm
- 38:39as you probably get, but you look at Germany, we thought
- 38:42about. Doing this in Paris, but it just
- 38:43seemed like too, too hard, like it was enough.
- 38:46Different in the continent, different language seemed
- 38:49challenging. But you look at the the LMU in
- 38:52Munich. I mean this is a lab where
- 38:54Stable Diffusion was invented. I mean, Stable Diffusion is at
- 38:57the base of pretty much all the image AI model and that's the
- 39:01discovery from, you know, from Germany.
- 39:04So I, I think Europe has a very, very strong voice, I think to
- 39:09play a lot to say in the AI world.
- 39:12And you are, are you betting just like quickly on local
- 39:16heroes or like are, you know, especially in the foundation
- 39:18model game, there's like the bet of, OK, the government basically
- 39:22is going to want some hometown hero and therefore there's an
- 39:25opportunity or is that not big enough for the most part?
- 39:27Well, I mean, I think there's defense, which is a very
- 39:30separate sector where I think there are some very European
- 39:33centric dynamics, which I think are, you know, are playing out
- 39:36and where the European market is big enough to create like big
- 39:40champion and a lot of value. I, I think if you look at Europe
- 39:44today, yes, I mean, you have foundational model, but if you
- 39:47look at what's being done in the rest of the world with, with,
- 39:50you know, 10s, hundreds of billions of dollar invested,
- 39:53we're not seeing this in Europe, but what we're seeing in Europe
- 39:56is great companies that are building on this model.
- 39:59I mean, you look at Synthesia, I mean Synthesia AI avatar, they
- 40:02are the leader globally. No one has done this.
- 40:05You have, you know, 11 labs, You have lovables like so Europe is
- 40:09capable of generating real, like global leaders.
- 40:13Yeah. And I think it's just the
- 40:14beginning of the cycle. And Philip, I would, I would
- 40:16actually include mistrial in that category and to to add to
- 40:19the French. List to be.
- 40:21Respectful that there is hard evidence that the regulated
- 40:25industries they play into specifically two of those
- 40:28banking and healthcare are siloed.
- 40:33And as much as we would love to sort of cheerlead those
- 40:35industries do have government regulation and they do have
- 40:38governments overseeing and and especially as as your AI moves
- 40:43more to the frontline. So in the case of Alan moving
- 40:46from back office to sort of processing tasks to all the way
- 40:51doing medical triage where the sensitive sort of HIPAA
- 40:55equivalent sort of patient data, you really have the regulator's
- 41:00breathing down their necks and saying this has to stay in
- 41:02Europe. For venture to work, we need
- 41:06exits. We've obviously got this sort of
- 41:09very enticing exit with scale in the news and clearly there are a
- 41:15lot of high salaries also being thrown around.
- 41:18But it feels, you know in venture overall there haven't
- 41:21there's never enough. We'd always like more of them.
- 41:25Yeah. I mean, how much can you I guess
- 41:27invest in companies on the back of these licensing Aqua higher
- 41:30deals? And then how, what do you see as
- 41:33sort of the best exit opportunities right now?
- 41:36And if if you're not conflicted, you can touch on the scale thing
- 41:40that some of you are, but I'm just throwing this out.
- 41:42I'll take a 15. Billion dollar whatever he likes
- 41:44and say, however. You call it deal any day.
- 41:47Is that well, is the 15 billion? That's really the money and the
- 41:50rest is like, maybe? I mean the rest is a stand alone
- 41:53company that's going to continue to operate with the CEO,
- 41:57customers etcetera. And then?
- 42:00Tom, how much is it worth as a stand alone?
- 42:03A lot less than 15. These guys invested in it.
- 42:07It was a phenomenal and is a phenomenal investment.
- 42:10And I think we've seen a new pattern of a sort of glorified
- 42:13Aqua hire. So to me, this is a sort of $15
- 42:17billion Aqua hire. We're seeing a lot of the
- 42:19customers move away for obvious reasons, like snorkel in our
- 42:23portfolios getting a lot of inbound because of it.
- 42:26But I actually respect Meta for this move.
- 42:29It's unbelievably, it's a critical time.
- 42:31You're as good or as bad as the team.
- 42:33They lost their head of AI, was it 3 weeks ago?
- 42:36And they've now brought in a world class operator that
- 42:38understands the secrets of, you know, LLMS and Gen.
- 42:42AI in Alexander Wang. And $15 billion is a lot of
- 42:46money. But what is it?
- 42:47It's 1% dilution or something. I would take that bet if I was
- 42:50that team. So I don't know how well the
- 42:53business will do on an ongoing basis.
- 42:55It might look a bit like Inflection did when they sold in
- 42:58a similar deal. Didn't hear so much from them
- 43:00afterwards. But phenomenal investment for
- 43:03these guys, $15 billion. You know, if you just wind back
- 43:06the clock five years ago and say $15 billion exit, it would have
- 43:11absolutely blown our minds. And now we're talking about it.
- 43:14We've just had a kind of recency bias of an open AI round where
- 43:17it doesn't sound like a lot of money.
- 43:19It was a phenomenal deal. Eric, let's not shop the future.
- 43:23Time will tell. Yeah.
- 43:24Time will tell. I love your if you're trying to
- 43:26create a rule, are we going to have more of those?
- 43:28We'll, we'll, we'll, we'll take your, your optimism, but you
- 43:33know, to more than agree with your, with the premise of the
- 43:35question. The economy does need some exits
- 43:40and the ecosystem needs some exits.
- 43:42And I mean, to take it even further, some sort of recycling
- 43:46of capital and moving on forward is the essence of realizing
- 43:50ambition for entrepreneurs and sort of, you know, putting the
- 43:54capital and talent back into into the startup ecosystem.
- 43:57So are we working hard to move our companies along?
- 44:01Yes, we are. But obviously a lot of the exit
- 44:06activity has has historically been on public markets.
- 44:09It's been challenging looking looking backwards.
- 44:13Let's hope it's more positive looking.
- 44:15Forward, Yeah. What's do you want US regulators
- 44:17to be chiller here or? Yeah.
- 44:20Is it, is it a regulatory problem with IP OS or you think
- 44:23it's a company quality problem? Well, it's it's been not too.
- 44:27We don't have the time to sort of analyze.
- 44:29That's a big question. But in in two sentences, the the
- 44:32bar has gone up in terms of marketing because of float
- 44:36research coverage, what's economical for the underwriters
- 44:39to bring to market. So in some sense the the pool of
- 44:44attention of both human time analysts by side as well as
- 44:50capital has shrunk and and as a result the bar has risen.
- 44:54And you know the the flip side of that is the activity in
- 44:58private markets. Yeah.
- 44:59Well, I think I would add the geopolitical uncertainty, which
- 45:02is a huge thing, right. I mean, you need markets, you
- 45:05need low volatility to go public and and right now I don't know
- 45:09what the volatility is because we have been living in high
- 45:11volatility for a very long time. Right.
- 45:13What about Felipe, like private to private and specifically like
- 45:17open AI? I mean, it's so valuable, but
- 45:19you're getting these weird shares like I don't know, do you
- 45:22think open AI can be part of the solution here and the and
- 45:24private to private deals? Databricks has certainly done a
- 45:27number of deals. I mean, I think private to
- 45:28private are one part of the the solution.
- 45:33But I do think we're going to, I think we need the public markets
- 45:37and the geopolitical environment to stabilize.
- 45:39Like hopefully there's signs that this is going to start to
- 45:43happen. I think when this happened, I
- 45:45think there's a very high number of high quality companies that
- 45:48are just waiting to go public and that's going to provide a
- 45:51good amount of liquidity in the system.
- 45:54I think for the strategic acquisition in AI, I think we're
- 45:56going to continue to see more because a lot of the big
- 45:59companies are seeing that they are missing these AI talents.
- 46:02And even if Meta is ready to some bold move and Meta is, I
- 46:05mean everybody says, oh, you know, they're behind, but
- 46:08they're still very good in the high, right.
- 46:11And so we'll see a lot of other companies saying, OK, now.
- 46:14We really need to clearly needs to make a move.
- 46:17Last question quickly. You know, technology is a boom
- 46:21and bust sort of business with hype cycles and corrections
- 46:25where anyone brave enough to sort of chart out or say where
- 46:28you think we are in this boom and bust cycle, How much do do
- 46:31we have a lot of juice over the next 12 months to keep investing
- 46:34aggressively on the AI thesis? Definitely an optimist on this
- 46:39one. I I go go back to the
- 46:41traditional industries, you know, fintech, banking, lots,
- 46:44lots to be done. Checkbooks open.
- 46:45You agree? Yeah.
- 46:46It's like the drivers for me are, is there a capital?
- 46:49There absolutely is. And if the IPO windows open,
- 46:51they'll be more in the private market.
- 46:52And then secondly, do we think that the models and the tech is
- 46:56getting better? And I think there's no end
- 46:57insight pre training and post training and then now with
- 47:01reasoning there's more progress. So right at our.
- 47:04Conference in November, we were all worried about a wall and
- 47:07then post training I think in particular showed with reasoning
- 47:09models there's plenty of improvement.
- 47:11Philippe last. Year 100%, yeah, I think the
- 47:13there there are two things. There is a cycle in prices which
- 47:16may go ups and down, but I think the secular shift with AI is
- 47:19here for the next 10 years. I think like we'll have a lot to
- 47:22invest in for the next 10 years plus.
- 47:24Great. All right, that's our panel for
- 47:25the morning. Riley is going to explain what's
- 47:27next.