Think You Know AI? 25 Startups Prove You Wrong

0:08 Welcome, welcome, everybody.

0:10 Should be a fun one.

0:12 Let me start with a question.

0:14 When I say AI, what's the first thing that comes to your mind?

0:18 For most of you in the audience,

0:20 it's probably your chat bot, or your favourite AI coding assistant,

0:23 or one of the many AI agents you have working on your productivity.

0:28 These days, it might even be OpenClaw or NemoClaw.

0:33 And these tools are great, right?

0:35 The sheer pace at which these tools have become

0:37 part of our everyday lives has been absolutely incredible.

0:40 They're always in the news and rightly so.

0:43 But today, in this session, we're going to talk about something else.

0:47 We're going to talk about AI in the physical world.

0:52 Thanks, Joshi.

0:53 So, you know, the trend of the 20th century,

0:56 right, was just this massive expansion of the world, right?

1:01 And in the 21st century, as we see AI start to bring us together,

1:05 start to connect more of that world,

1:08 we've seen its ability to interact with a huge array of problems,

1:13 right, to solve things that we didn't think were possible without it.

1:17 And so one of the things that we're gonna take you

1:18 through today is some major areas of the human experience, right?

1:22 Major areas of industry, of commerce, of health,

1:27 of other very deeply personal human aspects of the world.

1:32 And so I think what you'll see as we go through here

1:34 is that AI has not stopped in all of those areas, right?

1:39 And that hopefully this is some inspiration for what is possible

1:42 with AI and what startups are doing for real in the world.

1:49 So Chris just gave you a map of what we're gonna talk about today.

1:52 Let's make it more concrete.

1:55 We're gonna talk about 25 startups across six teams.

1:59 These startups are from all over the world.

2:02 They're at various funding stages, different founder backgrounds,

2:06 but all of them are solving hard problems in the physical world.

2:12 All of them are building on NVIDIA's tech stack,

2:16 whether it's at the edge or in the cloud.

2:20 Now, we have 25 startups to cover in roughly 30-ish minutes,

2:25 so things are going to move fast, no doubt.

2:28 You can think of this more as a guided tour of what

2:32 these startups are working on to build awareness than a proper deep dive.

2:36 So with that, let's get started.

2:39 Let's talk about agriculture.

2:43 Agriculture is one of the most important and oldest human industries.

2:47 In the 20th century, We produced more food than ever before.

2:54 As our population grew enormously, agriculture expanded too.

2:59 But that expansion came with challenges.

3:03 Soil depletion, heavy chemical use, water wastage.

3:08 And now, in the 21st century,

3:10 as our population has grown to 8 billion, we deal with the same challenges.

3:16 In fact, The World Food Programme says that AI and ML

3:22 offer unprecedented opportunities to accelerate

3:25 our progress towards achieving zero hunger.

3:28 So in the 21st century, as our population grows even more,

3:33 how do we feed 10 billion people while also solving world hunger?

3:38 And at the same time, mitigating the challenges I just talked about.

3:42 Let's see how AI can help us do that.

3:46 So I love this one.

3:47 Look, you know, we talk about deep learning all the time,

3:50 right, to the foundation of modern AI.

3:51 This is gonna be shallow learning today because that's

3:53 all the time we got for each of these startups.

3:57 I love this kind of the mission of this startup, right?

3:59 I didn't think about, before meeting this mission here,

4:05 I hadn't really thought, frankly,

4:07 about what it takes to make sure that pollinators can have healthy lives,

4:12 right, healthy colonies.

4:13 Some of you have maybe heard of the big

4:16 challenges in colony collapse among pollinators in agriculture, right?

4:19 And so it turns out that one solution to that is to give bees luxury apartments.

4:27 And that's barely even a joke, right?

4:29 This is, these boxes that you can see here are fundamentally

4:33 the kind of automated system for holistic health management for bees.

4:40 Using edge computing, computer vision,

4:42 using all kinds of sensors to determine if the bees are happy,

4:47 if they're healthy, if they've got parasites.

4:50 And being able to have that kind of always-on

4:54 monitoring enables a much larger, to your point,

4:59 much larger coverage of the agriculture than if we had to, you know,

5:04 humans drive out and inspect every one individually.

5:07 So you can have that kind of persistent measurement of what's going on there.

5:12 So happy bees for our agriculture.

5:16 You're going to see in the next two slides here a couple

5:18 of different ways that robots are playing a major role in advanced agriculture.

5:24 So this company in particular, by doing edge computing,

5:28 we spend a lot of time these days, like you said,

5:31 Joshi, thinking about cloud and really massive,

5:33 that was a big subject That's the concept of Jensen's keynote.

5:37 is how big of scale we can go.

5:40 We're also kind of thinking at NVIDIA about how small can we go, right?

5:43 How much can we put at the edge with embedded compute?

5:47 And so by having the compute at the edge flexible enough, right?

5:52 Not just single function chips,

5:54 but flexible enough because you have GPUs to be able

5:57 to then adjust to different kinds of model depending on the need.

6:01 Make sense?

6:03 So that captures the traditional methods for agriculture,

6:06 weeding and spraying, right?

6:07 But also, again, thinking of everything as a sensor, right?

6:10 Thinking of everything as a way to continually

6:13 improve and grow the foundations of agriculture.

6:19 And in a particular kind of sense, right, one of the challenges, like we said,

6:22 is how do we provide nutrition for the billions of people that we have to?

6:27 And if we were stuck waiting on agriculture that is

6:32 appropriate to the climate of everywhere individually in the world,

6:36 that would be a much bigger challenge.

6:38 Here we're seeing with greenhouses, and particularly with automatic greenhouses,

6:42 robot-assisted greenhouses, we can grow healthier,

6:46 more robust food In places that it might not be able to otherwise.

6:49 And so that's a huge change, kind of a step change,

6:51 in our ability to provide a variety of nutrition

6:54 all the time for populations around the world.

7:00 And then let's zoom in on a really

7:02 specific part of the production chain here, right?

7:05 So sustainability, like you said, Joshi,

7:07 is a huge part of how we're gonna feed 10 billion people.

7:11 And so thinking beyond just kind of the normal weeding and spraying,

7:15 thinking to how do we do that without So,

7:19 on the left you can see here a company that's using mechanical hose,

7:24 slamming this kind of metal implement into the dirt and pulling out

7:28 the weeds as it's going by by using vision at the edge,

7:32 right, to detect when there needs to be

7:34 a weed literally physically pulled out of the ground.

7:36 Solar-powered, actually even wind-powered and using all of that technology

7:41 to become kind of autonomous at pulling out these weeds.

7:45 On the other hand, on the right side here,

7:47 you can see what happens when we put lasers on similar machines, right?

7:50 When you can have, at the edge, micro-focused lasers killing weeds with energy,

7:55 right, instead of with chemicals.

7:57 Pretty cool to see how that all enables

8:00 scale in a way that wasn't possible before.

8:02 Yeah, same problem, different methods.

8:04 Absolutely.

8:06 We just talked about farming and agriculture.

8:09 Let's talk about protecting communities in the world.

8:13 Think about the last wildfire you saw in the news.

8:18 By the time it made the headlines and people found out, it was too late.

8:23 A lot of damage had already been done.

8:24 Neighborhoods were wiped out.

8:26 And think about the world we live in today, right?

8:29 The population is increasing, our cities are getting more dense and complex,

8:34 the weather's unpredictable,

8:35 and so we're at risk more so than ever for these threats.

8:39 And it's not just wildfires,

8:42 it's storms, floods, other urban emergencies, right?

8:48 These are not rare events anymore, right?

8:50 In fact, the UNEP says that extreme wildfires are projected

8:54 to increase by up to 50% by the end of the century.

8:58 Think about that, up to 50%.

9:00 So these problems are gonna get harder, not easier.

9:05 Now, in most cases, The early signs are there.

9:09 We just don't detect them early enough to do something about them.

9:12 And therein lies the challenge, right?

9:14 How do we shorten the gap between when one

9:16 of these threat occurs and we can do something about it?

9:20 Let's see how AI helps us do that.

9:23 So it turns out that there are signals

9:26 of potential wildfires in lots of different places.

9:29 In some cases, the signal is very much at the ground level,

9:33 especially the micro signals,

9:34 being able to see kind of individual patches of a forest burning.

9:39 Or starting to burn, right?

9:41 And so that's what we're seeing on the left here,

9:42 is how do we put ground-based sensors empowered by compute at the edge,

9:46 empowered by models that are driven by training, not at the edge, right?

9:51 And then use that to shorten the time between occurrence of some threat, right?

9:57 And detection of that threat.

9:59 Be able to get out ahead of it.

10:01 And it turns out that the other direction here is that one,

10:04 if you think of one sensor as looking horizontally,

10:06 like looking out over a forest,

10:08 over can be a very different word if we say over in a satellite sense.

10:13 And so it turns out you can put GPUs on satellites.

10:16 It's a pretty good place to do your advanced computing

10:19 so you're not beaming down all of your images at scale,

10:22 these huge images that if you want to cover huge tracts of land.

10:25 It's much more efficient if you can process

10:27 it on board the satellite and beam down

10:30 just the detections or the zoomed in parts of what you need to be looking at.

10:35 So having these two very different approaches

10:38 to what is fundamentally a major similar problem.

10:42 Is a huge part of conceptually how we should think

10:45 about what startups are able to bring to AI, right?

10:47 That it's not about one solution to a problem, right?

10:50 It's kind of, it's about a thousand solutions to the problem.

10:53 And then figuring out which of them is most successful

10:55 in doing the thing that we need it to do.

10:59 There's a kind of interesting follow-up to that, right?

11:01 If you think of the left side here,

11:02 maybe even the right side as adding new sensors, right, into the world.

11:07 I love when startups, you know, have a major part of their business

11:12 being collecting more data about the physical world.

11:15 And so that's where, you know,

11:16 it's not enough just to have kind of a sensor, right,

11:19 or to rely on legacy sensors that weren't,

11:22 we should remember, weren't laid out for AI processing, right?

11:26 Traditional weather sensors weren't laid out for how

11:29 do I maximize the fidelity of my AI models.

11:32 But if you start from the premise that I need to make as good

11:36 of a detection chance as I can have for a particular kind of AI,

11:39 then putting up radars might make sense for you.

11:42 That's what we're seeing on the left with climate vision here,

11:45 literally putting up a radar network kind of across

11:48 the south-southeastern United States that then helps augment

11:53 places where otherwise traditional weather forecasting methods may

11:56 have had gaps or been less able to predict.

12:00 The kinds of disasters that threaten

12:03 particularly kind of heavy population centers, right?

12:07 The kind of things that, you know, have caused so much damage as extreme climate

12:12 has proliferated as you described from the UN.

12:16 Another way to think of this, though, is to go from sensor processing.

12:19 How do we, instead of just taking images or from traditional sensors,

12:22 how do we really get a detailed picture?

12:25 In some sense, this is about how do we shape our understanding of the world?

12:30 The other side of this, on my right here, is how do we,

12:33 instead of going from traffic cameras to a 3D

12:37 understanding of what's happening in our cities?

12:40 To be able to say, hey, this problem isn't just about counting cars but about

12:45 being able to track flows of people and of things

12:49 through cities so that we can treat them as integrated

12:52 systems rather than as individual points on a map.

12:57 And so it's really cool to think about it from, again,

13:00 two very different angles on a similar problem, right?

13:02 If we have challenges affecting our cities,

13:05 adding new sensors and interpreting those sensors in different ways

13:09 really provides some powerful leverage to adapt to those problems.

13:17 So we just talked about protecting communities,

13:19 and the key there was early detection, right?

13:22 The theme was urgency.

13:24 Let's talk about the environment where

13:27 problems build up gradually, slowly, right?

13:30 When you think about waste management systems, recycling, water systems,

13:34 forests, the problems in these areas are slow and gradual,

13:38 and we don't often find out about them until it's too late, right?

13:44 Let's go to the board.

13:47 But the problem is that if we cannot see these problems,

13:51 we cannot measure them, right?

13:53 We don't do a good job measuring them.

13:55 And as the UN says, 68% of environment-related

13:59 sustainable development goals lack sufficient data to assess progress.

14:04 Let's think about that.

14:05 We've set our goals, for 68% of them,

14:08 we don't even have the data to measure and see if we're making progress, right?

14:13 So we're just flying blind.

14:15 And so the opportunity here is how can we

14:18 use AI to get more visibility into the system?

14:22 So one, we can start measuring them, and then two,

14:24 we can start tracking and see if we're making progress against our goals.

14:29 Let's look at a few examples.

14:31 So look, it is very easy to get excited about the shiniest version of AI, right?

14:37 The version of AI that talks about building

14:39 the sci-fi glass and steel version of the future, right?

14:44 But I'll be honest, I get just as excited about the really low level micro,

14:48 you know, it's there and we try not to think about it, parts of our world.

14:54 And that's what we're seeing here with these companies.

14:56 They're dealing with our waste streams.

14:58 In some sense, the challenges that we're solving by having denser cities,

15:02 smaller transportation costs, more interconnected communities,

15:06 they come at a different kind of cost.

15:09 Now we have to deal with the waste or the recycling

15:12 that we need to do from those human populations.

15:15 And so if we can instead apply AI to Analyzing,

15:19 getting data on, as you can see on the left here,

15:22 scanning to understand what's in our waste flows.

15:27 So that instead of just saying there is a lot of it,

15:30 we can do something with that.

15:32 We can turn this into the kind of optimizable

15:35 cycle that we're so used to in tech and bring

15:38 that to industries that have lots and lots of potential

15:41 for really applying 21st century tools to these problems.

15:47 So once you have the data, just like you were saying,

15:49 once we have the data, then we have the ability to make progress on these goals.

15:52 And so that lets us then, for instance,

15:55 add robots into the process to help us sort and pick from those waste streams,

15:59 to bring out of those waste streams things

16:02 that don't have to be there, again, at scale.

16:07 Right, not just, it's not just kind of another dirty job.

16:09 This is something that now we can kind of solve problems at city scales,

16:13 which is not trivial, but it's exciting that, you know,

16:17 startups are taking that on.

16:18 And it's really exciting to me personally,

16:20 that the market has appreciated that, right?

16:23 That these startups are able to access the kinds of people

16:28 and technologies that they need to make

16:30 this meaningful contribution to the world.

16:33 If we go kind of hard the other direction, you think about,

16:35 okay, that was, that's thinking about kind of urban environments, right?

16:38 If we're reading the signals of the planet on a more geophysical scale, right?

16:44 Looking at systems that again, you know,

16:46 on this theme of what sensors do we need?

16:48 To tell if we are making progress, and if we are,

16:52 how much progress towards our macro goals as a society.

16:56 Putting up weather balloons, again, it's not the kind of thing that's like,

17:01 you know, making the shiniest possible headlines.

17:04 It's super important.

17:06 Right, otherwise we can do all of the things that we want in the world,

17:09 but we don't know if they're having the real impact.

17:12 And I care a great deal about that.

17:15 The other side of this, if we think of collection as the left side,

17:19 we can think of on the right,

17:21 how do we make the data that we collect accessible,

17:24 a common operating picture of the planet?

17:29 But that's not easy.

17:30 There have been lots of cases where people have started trying,

17:33 but bringing that together in a way that we can have,

17:36 you know, that people will have

17:38 a common operating picture of their business, right,

17:40 or an industry, if we take that same mindset and we're able to say, hey,

17:44 it's just as important to have a common understanding of the planet,

17:47 that we bring data together, make that data available to...

17:51 Consumers of it downstream in a way that has

17:54 not necessarily been available and to use AI to both

17:58 accelerate and improve the precision of that kind of processing

18:02 so that then it is available at planet scale.

18:08 Let's talk about something that everything else depends on, the power grid.

18:13 Now, most of us don't think about the power grid until our lights go out, right?

18:19 But it is something that everything else depends on, hospitals,

18:23 transportation systems, communication, and of course, data centers.

18:30 But the grid was built decades ago.

18:32 It was built for another world.

18:34 It was built for a world where power flew in one direction,

18:38 from the utility plant to people's homes and businesses.

18:42 Think about how that's changed now.

18:44 We now have solar rooftops pushing electricity back into the grid.

18:48 We have EVs charging and discharging rapidly at unpredictable times.

18:55 Now all of this, plus the fact that one, the grid is old, and two,

18:58 it has to deal with extreme weather means

19:01 that there is more strain than ever on our grid.

19:06 And inevitably, when the grid fails, not everyone is impacted equally.

19:13 The Council on Foreign Relations says that when severe storms strike,

19:17 lower income communities wait the longest for power to return.

19:20 Just think about that.

19:22 The people who need power the most have to wait the longest to get it back.

19:28 Now, one solution is to just rebuild the grid from scratch,

19:32 but that's not practical.

19:34 So how can AI help us make our grid more resilient, flexible, and smart?

19:41 Let's look at a few examples.

19:42 Yeah, absolutely.

19:44 You know, look, one of the challenges of living

19:47 in an environment where we are sharing the planet, thankfully,

19:50 with lots of plants is that they don't always have

19:54 a purely positive impact on your access to electricity, right?

19:59 If you've had a tree or branch fall on power lines,

20:04 you've certainly experienced what I'm talking about.

20:06 And so it turns out that there are,

20:09 of course, startups dealing with those problems too.

20:13 Here, the trick is satellite imagery, right?

20:17 Being able to look down

20:19 at a fine-grained understanding of the local environments,

20:22 figure out that there is foliage putting at risk parts of the grid.

20:28 And then be able to proactively address

20:29 that in a way that so far has very much been,

20:33 like you said, unevenly distributed, you know,

20:35 human-based patrols looking for those kinds of risks.

20:39 Being able to do that again at scale all

20:41 the time on kind of a persistent defense strategy

20:43 is a huge part of how we make grids

20:47 more resilient with the systems that we already have,

20:49 right, without doing a wholesale kind of ground-up revision of the grids.

20:57 If one version of this is my usual story,

20:59 hopefully you don't get too tired of we need more sensors,

21:02 more data processing, and then more action based on the data,

21:05 that's kind of the theme of how AI works.

21:08 Here's a specific kind,

21:09 a flavor that we haven't talked about yet but is very cool

21:13 and is something that NVIDIA has played a significant role in, digital twins.

21:18 Being able to have in the computer a kind of physically driven,

21:21 sensor-informed model of how the grid works,

21:25 being able to use that model to simulate different kinds of stress on it,

21:29 gives us the ability to, again,

21:32 proactively kind of move forward the action point and not just

21:36 be waiting for it to fail and responding to that, right?

21:39 Because if we do the latter, again, you get these kind of inequities in how...

21:44 in how those resources are applied, where if we can get out ahead of things,

21:48 we have the ability to prevent rather than react.

21:53 And then, you know, like Joshi said, right, if we say,

21:55 hey, the grid was for 100 years, 150 years,

21:59 has been electricity coming from generation sources and flowing out to people,

22:03 right, if we invert that paradigm and we think,

22:06 hey, there are lots of generation sources now,

22:08 there are lots of, we kind of have to do this two-way street,

22:11 and that just, it doesn't just, you know,

22:13 increase the complexity, right, it kind of squares the complexity.

22:17 Because now everybody is a consumer and a producer

22:19 in this simplified model of the world.

22:22 And so now you've got all of the flows going from everywhere to everywhere.

22:26 It's much more complicated.

22:27 And so it doesn't really work with the old models.

22:30 And so then having particularly sensors at the edge that let us

22:34 have an always-on understanding of the whole back and forth of the grid.

22:40 As we're preparing for data centers and the impact, as you described,

22:44 that they're having on the environment, to use AI to mitigate the impact of AI.

22:51 And that's what we're seeing on the right side here,

22:54 that the only, in some sense,

22:56 the only solution that we're going to have for how do we create a sustainable

23:03 growth pattern in these industrial areas is going to be the kind of modeling,

23:08 the kind of data analytics, the kind of forecasting, the kind of simulation,

23:13 and then building around that that is made possible by AI.

23:19 For the new grid, for a grid that is people-centric,

23:23 you have to really change the origins of where you're thinking.

23:29 All right, we're halfway there.

23:32 So far, we've been talking about big systems, like planet-level stuff.

23:35 We've talked about farms, we've talked about cities,

23:38 talked about the environment, we've talked about the power grid, of course.

23:43 Let's zoom all the way in down to an individual.

23:46 Now, some of the most important moments in people's

23:49 lives also happen to be the most private and quiet.

23:53 Let me give you an example.

23:55 There's probably an elderly person right now that might have a fall.

23:59 But because they live alone, it might be hours before someone finds out, right?

24:05 Because these moments are private and quiet, not everybody's watching, right?

24:08 Not everyone has access to a caregiver.

24:11 Not everyone has access to a specialist.

24:14 Not everyone has someone or something watching them

24:17 at those important moments when they need it the most.

24:23 And I think this is where AI can help, right?

24:25 The ITU Focus Group says that AI-based

24:28 technologies hold great potential in improving the accessibility,

24:32 quality, and value of healthcare outcomes.

24:35 In this case, this is as much an accessibility

24:39 problem as much as a tech problem, right?

24:41 How can AI help us reach more people in the moments they need it the most?

24:47 Let's look at a few examples.

24:50 One of the biggest challenges worldwide in health is vision impairment.

24:56 There are low cost interventions, right?

24:59 Glasses that still need to be proliferated around the world.

25:03 There are sophisticated interventions, right?

25:06 Dogs as companions for visually impaired folks,

25:09 but those are frankly quite expensive and very unevenly distributed.

25:15 And so if we can bring down the cost,

25:18 if we can bring in the learning from people around

25:20 the world as to what they need to navigate their environments,

25:23 we have the ability now to build technologies that, like this one,

25:29 give freedom back to people who need it, right?

25:32 This headset here has sensors, onboard processing, again, that...

25:39 Let you see without sight, right?

25:40 They'll let you see in the same way that cars

25:43 do and so you can you can see the the kind

25:45 of demonstration here is being Able to reach out and grab

25:48 an apple With no visual data going to your eyes.

25:52 That's, it's remarkable.

25:54 And again, I think the thing that I'd hopefully

25:56 leave you with here is that the, making that available, making that accessible,

26:00 making that widely distributed, right, is in some sense the thing that is most

26:05 important to me about how these technologies are changing lives.

26:09 It's less directly about how cool it is, and it is very cool,

26:13 but it's really about how this can be replicated

26:16 more easily than many of the alternatives have been historically.

26:20 If we think, you know, I kind of heard the other direction,

26:23 right, if this is let's make everybody's day to day more accessible,

26:27 right, in the worst and the hardest and the most stressful moments,

26:31 like you're saying, we also need people watching our backs.

26:35 And emergency first responders need people watching

26:38 their backs as they try to provide care under,

26:41 again, some of the hardest times that you can experience.

26:44 And the last thing we want to do is have people,

26:47 you know, making their best judgments under,

26:49 you know, significant stress and having to monitor

26:52 lots of different things all at the same time.

26:55 And so one of the things that AI can do, it turns out,

26:57 is understand how, what people are hearing on emergency dispatcher calls,

27:03 right, 911 calls here in the U.S.,

27:06 how people's voice patterns and stresses and other

27:09 signals can tell us about their individual situation.

27:12 Right, and take off some of that cognitive

27:14 load from the dispatchers and the clinicians

27:16 who are being asked to provide diagnoses

27:18 and response while still keeping them in charge,

27:22 right, acting as kind of a supplement

27:24 to their expert judgment rather than something else.

27:28 We also, as Josie said, in kind of the quietest moments,

27:31 in the moments where you least want people constantly monitoring,

27:36 to have AI be able to provide that impersonal privacy-protecting support.

27:44 Being able to say, hey, this senior needs help,

27:48 without having to have a person monitoring that.

27:51 To be able to understand early in your vocal patterns,

27:54 if there's something cognitively that is impairing

27:57 or an early sign of something happening in your body,

28:02 incredibly important as we're going to deal with the challenges

28:06 of aging populations and in equal access to healthcare.

28:11 And so again, in all of these cases,

28:13 the AI is serving the role of spreading and of making broader

28:18 the kinds of health care that otherwise would be expensive and limited.

28:25 I'm excited for that world,

28:26 and I know a lot of people very close to me who are benefiting that way.

28:32 Right, in all the sections we've covered so far,

28:37 the team has been seeing things early and more clearly.

28:42 That's true in this one as well.

28:43 Let's talk about science and medicine,

28:45 where some of the most important decisions also happen at the very beginning.

28:51 Now, we have the most amazing people working in science,

28:54 but science itself has gotten too complex.

28:59 200 years ago, you could argue that one smart individual could

29:03 probably keep track with everything going on in the field of science.

29:07 But that's just not true anymore.

29:09 We've made so much progress, and there's so much information out there, right?

29:14 It's impossible for one person to keep track.

29:16 We now have thousands of teams working on tens of thousands of projects,

29:20 all solving small parts of the big puzzle that is science.

29:26 Now, if you think about it, before a trial or a treatment begins,

29:29 these smart people have to make some early but consequential decisions.

29:34 Which path to take for a given trial?

29:36 Which direction to take for a treatment?

29:39 And these early decisions ends up shaping years of research after.

29:44 The challenge is that these folks are making

29:47 these decisions often without having the full picture.

29:50 Therein lies the opportunity, right?

29:52 AI is enabling the analysis of extensive

29:55 data sets and also helping uncover hidden patterns.

29:59 So if you think about it, if we can augment our amazing

30:02 scientists and researchers with the complete information,

30:05 if we can augment their intuition with the full picture,

30:08 right at the beginning when those decisions really matter,

30:10 then we can accelerate the pace of discovery.

30:14 Let's look at a couple of AI startups that are helping us do that.

30:18 You know, like you said Joshi, it's really challenging, right,

30:21 when you have to pick paths without being able to, for very good,

30:28 in this case, privacy protective reasons, right,

30:31 that aggregating kind of all of the healthcare data of patients

30:35 with a particular condition or a particular family of research,

30:38 that would be enormously helpful to the model, but a significant risk.

30:43 And so, one of the things that is, I think,

30:45 underappreciated about the modern kind of deep

30:48 learning training solutions is what we,

30:50 you'll forgive the jargon here, but federated learning,

30:53 where I can take a model and move it to the hospital, right?

30:57 Move it to where the patient data is stored,

30:59 rather than moving the data to where the model training is happening, right?

31:03 And so that means that every medical facility,

31:05 all of these protected enclaves of data,

31:08 the model comes in, gets a little bit of training, and then moves on.

31:13 And then if you do that, turns out you're training on the data.

31:16 You are still building the model.

31:18 You're just doing that in a way that then

31:21 doesn't aggregate everybody's medical and health information to one place.

31:25 So that's really exciting, right?

31:26 You can facilitate, you can power a whole new generation of models without

31:32 having the kind of risks and concerns that you might have had otherwise.

31:36 But then you can build on top of that, right?

31:39 You can say, okay, we're going to simulate We're going to, just

31:44 like we were with the kind of digital twins for the power grid,

31:48 we can simulate people's biology

31:50 and, as the caption here observes, test drive, right?

31:55 Try out a treatment in this model, in silicon, right?

32:00 In a way that then gets to that high

32:03 leverage point at the very beginning of treatment.

32:06 Or if we're designing new ways to test, as you can see on the right here,

32:12 design new ways to test and pipeline

32:14 through interventions without having to limit

32:17 ourselves to traditional ways of assessing the data that you're getting, right?

32:21 Be able to test them on realistic surrogates, analogs,

32:25 for the actual place that you want to deploy a therapy.

32:30 And so being able to, again,

32:31 create from the ground up these very novel places to make the decisions,

32:37 assisted by AI, that's an incredibly powerful way to do medicine.

32:42 I'm excited for that.

32:44 And then here I'll give you kind of the In some sense,

32:51 if I've got drugs, pharmaceuticals, and I've got my genes,

32:55 the hope is that I can have some

32:58 kind of personalized therapy that goes along with that.

33:01 If you think of going from drugs to outcomes,

33:04 being able to inject the response that my particular

33:07 genetic makeup will have to those therapies is incredibly powerful.

33:11 And it's something that, when we talk about personalized medicine,

33:14 right, there are lots of ways to personalize medicine,

33:16 but one of the most fundamental is

33:19 to understand how I might personally respond to something,

33:22 to a particular treatment.

33:24 Again, being able to intervene at that high leverage point early

33:27 on and not have to wait until I'm seeing the reaction as a patient.

33:36 All right, so everything you saw here today is about

33:41 how AI is helping solve hard problems in the physical world.

33:47 We just covered a lot.

33:48 We just covered 25 startups across six teams.

33:53 Now, I don't wanna recap everything,

33:54 but I did wanna leave you with a few, right?

33:58 You saw AI in farming, how it is helping farmers treat at the individual

34:02 plant level instead of the farm level.

34:05 We saw satellites detecting wildfires from space in a matter of minutes.

34:10 We saw robots helping sort recyclables so less of it ends up in landfill.

34:17 We saw how AI can help us create a digital twin of our power

34:20 grid so we can spot weaknesses and be better prepared for emergencies.

34:26 We saw a pair of glasses that can

34:28 help the visually impaired navigate the world independently.

34:33 We saw AI can help us create a digital tumor to help doctors test

34:38 treatments or test which treatment would work

34:41 for a patient before the actual treatment began.

34:43 And like I said before, all of these companies are solving hard problems

34:49 and all of them are building an NVIDIA stack.

34:53 The other thing that's common for all 25

34:55 is they're all part of NVIDIA's inception program.

34:59 It's how we support these startups by giving them access to free compute,

35:10 technical resources, go-to-market support,

35:12 connecting them with the broader NVIDIA ecosystem.

35:16 We help them from zero to one and then one to 100.

35:18 We are there throughout the way.

35:21 Best of all, it's free.

35:24 We work with startups in all industries,

35:26 Including the non-obvious one, as you saw today.

35:30 And we have startups from over 120 countries that are in Inception.

35:35 In fact, some of the startups that you saw today are here at GTC.

35:38 If you wanna learn more about Inception or wanna

35:41 learn more about the startups that we talked about today,

35:44 please come talk to us at the Inception info desk in the expo hall.

35:48 And then lastly, if you're building a startup

35:51 or you know someone that's building a startup, please join Inception.

35:55 The link is right there.

35:57 We'd love to work with you and be supportive of your journey.

36:01 Thank you.

36:02 Thank you so much.