Latest / Automated with Brian Heater / Dr. Ahmad Bahai on the Semiconductor Race Behind AI and Robotics

Transcript

Dr. Ahmad Bahai on the Semiconductor Race Behind AI and Robotics · Automated with Brian Heater

• 0:00
  SPEAKER_01: I believe that there is no other technology in human history as creative and as innovative as semiconductor.
• 0:09
  SPEAKER_01: Sometimes scientists say that, okay, here is the limit.
• 0:11
  SPEAKER_01: You cannot go beyond that limit because it's built on some physics law.
• 0:15
  SPEAKER_01: And it's amazing how often I've seen engineers surprise scientists to find a way to go around that.
• 0:21
  SPEAKER_00: As I started covering AI and robotics, one of the things that's very interesting to hear, very technical people talk about things like magic.
• 0:30
  SPEAKER_00: You've got the people who are working on them that are completely surprised that they work the way they do.
• 0:36
  SPEAKER_01: 10-15 years ago, there was a very famous physicist who said, okay, you cannot go below 55 nanometers.
• 0:41
  SPEAKER_01: And guess what?
• 0:42
  SPEAKER_01: We are now at 2 nanometers today.
• 0:45
  SPEAKER_00: In the past 10 to 15 years, what if anything has the company been doing to prepare itself for this moment?
• 0:52
  SPEAKER_01: Nobody can predict the next market in 10 years.
• 0:55
  SPEAKER_01: But what you can see is that the trends are very clearly there.
• 0:59
  SPEAKER_01: For example, now that we have these massive data centers, one of the bottlenecks is how to deliver power to these giant chips.
• 1:06
  SPEAKER_01: Demand for power from portable devices all the way to data centers is insatiable.
• 1:11
  SPEAKER_00: Is there a way to have more and more and more compute demands to see that this is going to be more and more research intensive?
• 1:19
  SPEAKER_00: To see that at some point, like, you know, just not get more doom and gloom about it.
• 1:42
  SPEAKER_00: Big guest this week, we are speaking with the Texas Instruments Chief Technology Officer, Ahmad Baha'i.
• 1:49
  SPEAKER_00: We'll dig into his backstory, TI's rich history, and how the chipmaker is working to remain competitive in the era of physical AI.
• 1:57
  SPEAKER_00: Thanks to Ahmad, and thanks to you as always for tuning in.
• 2:00
  SPEAKER_00: If you've been enjoying the podcast, the easiest way to support us is by liking, subscribing, telling a friend.
• 2:06
  SPEAKER_00: I would also encourage that you check out our newsletter of the same name.
• 2:10
  SPEAKER_00: You can find that over at automated.fm.
• 2:12
  SPEAKER_00: For now, stick around and enjoy this conversation with Ahmad.
• 2:16
  SPEAKER_00: Bye.
• 2:17
  SPEAKER_00: Bell Labs is one of those legendary places.
• 2:21
  SPEAKER_00: You were there for a minute.
• 2:22
  SPEAKER_00: I'm I'm very curious what it was like in the 90s.
• 2:27
  SPEAKER_01: Yeah.
• 2:27
  SPEAKER_01: Hello, Brian.
• 2:28
  SPEAKER_01: Yeah, absolutely.
• 2:29
  SPEAKER_01: It was quite an experience as a fresh graduate PhD student.
• 2:33
  SPEAKER_01: Uh going to Bell Labs was being uh in the most exciting research institute in the world.
• 2:40
  SPEAKER_01: Uh in fact, uh when I finished my PhD, I was thinking about should I go to academia, to industry, or go to Bell Labs.
• 2:47
  SPEAKER_01: And I was at some point convinced that Bell Labs is the place because you walk in the place and you see the giants that you really know them from, the textbooks that they wrote for and a Nobel Prize that they uh which they won, and all the presence they had in the entire industry.
• 3:05
  SPEAKER_01: So I was fortunate enough that when I joined Bell Labs, I worked with one of the Nobel laureates that is Bell Labs.
• 3:12
  SPEAKER_01: And Bell Labs had more Nobel laureates than almost every university out there.
• 3:16
  SPEAKER_01: So that kind of mentorship for a fresh graduate was quite an exciting experience.
• 3:23
  SPEAKER_01: I don't think there is any place like that anymore in the world with the concentration of talent and the freedom of exploration because of the unique situation of Bell Labs.
• 3:34
  SPEAKER_01: Pretty much as long as you can convince your technical leader that the topic is worthy of doing research, you had a chance to work on it.
• 3:42
  SPEAKER_01: And that could range from theoretical physics, material science, electronics, information theory, and everything in between.
• 3:51
  SPEAKER_01: And that environment was very stimulating, very inspiring, and very important.
• 3:56
  SPEAKER_01: And I really enjoyed the time.
• 3:58
  SPEAKER_01: I was there for four or five years, and that was uh an amazing uh experience that really built the foundation for everything that I did since then.
• 4:06
  SPEAKER_00: What was the focus that you ultimately settled on?
• 4:10
  SPEAKER_01: So originally I started with the wireless communication and satellite communication.
• 4:16
  SPEAKER_01: So I worked on very early stage of uh uh Wi-Fi on concept back then in the 90s.
• 4:23
  SPEAKER_01: And uh interestingly, after two years, three years there, I wrote a book on the topic of multi-carrier OFTM systems, which is really the foundation for Wi-Fi and 5G and LTEs and a lot of other technologies.
• 4:38
  SPEAKER_01: And working with people who pretty much started field of communication many decades back was such an amazing experience to work with them, understand some of the uh learnings that they had, some of the insights that they can share with you.
• 4:53
  SPEAKER_01: So, yeah, I worked on wireless systems.
• 4:56
  SPEAKER_01: I remember working on some of the early advanced basis stations and some of the technologies for um wireless local area network, what we call it Wi-Fi today.
• 5:07
  SPEAKER_01: So that was the and also on devices like power amplifiers and new um materials that can offer higher performance in in wireless communications.
• 5:18
  SPEAKER_01: It was kind of a mix of system device and circuit that was really unique to have.
• 5:24
  SPEAKER_01: And as I said, in each field, you have men you had mentors that you really could connect with and and interact with and enjoy the intellectual discourse.
• 5:34
  SPEAKER_00: Yeah, I I want to get back to this idea of of graduating and and having these three paths ahead of you.
• 5:41
  SPEAKER_00: Um because we talk about this a lot on the show, and and it sounds like ultimately in your career you've been able to do uh a little bit of both, right?
• 5:50
  SPEAKER_00: And you know, you you're you've been working in this corporate environment, but obviously, you know, you you remain involved at places like uh MIT, for example.
• 5:59
  SPEAKER_00: But but what you where you started career your career was sort of this this middle ground, right?
• 6:08
  SPEAKER_00: I mean, Bell Labs is sort of this midway point between the corporate world and the pure sort of academic research world.
• 6:15
  SPEAKER_01: Yeah, and it was a unique place in that sense.
• 6:18
  SPEAKER_01: Uh there are not that many places like uh Bell Labs anymore because of the unique setup that Bell Labs and ATT had back then.
• 6:26
  SPEAKER_01: But you're absolutely right.
• 6:27
  SPEAKER_01: Bell Labs was the best of both worlds.
• 6:30
  SPEAKER_01: I mean, you had one foot in theory and academic research.
• 6:34
  SPEAKER_01: I mean, you had some of the best researchers in the field, and on the other hand, you tie with the applications that really what makes those research impactful in the real world.
• 6:43
  SPEAKER_01: So it was a very unique setup that really combined the depth of the research with the reality of what it takes to make them uh make these systems impactful and and and consequential.
• 6:58
  SPEAKER_00: Are you trying to foster a somewhat similar environment with Kilby Labs?
• 7:03
  SPEAKER_01: That's exactly what I tried to do in a smaller scale.
• 7:05
  SPEAKER_01: Of course, Bell Labs was the place with an amazing uh political mass of people.
• 7:11
  SPEAKER_01: But at the personal level, since uh uh I graduated from Bell Labs, I mean, after many several years there, I always felt that having one foot in academia and one foot in industry can help you bridge that gap.
• 7:24
  SPEAKER_01: I mean, the understanding of what the theory says and what the science and uh leading-age technology offers and how to implement it, I found this as an important factor that is missing from a lot of research institutes.
• 7:39
  SPEAKER_01: Of course, uh, the size of Kilby Lab is much smaller than what Bell Labs would have been like because of this special situation of Bell Lab since its uh uh inception early 1900s.
• 7:51
  SPEAKER_01: But uh the idea is that exactly how to interface between advanced research that is done in university but may not be manufacturable, and what you need in industry and how to bridge that gap.
• 8:04
  SPEAKER_00: Yeah, I leave, you know, we've had people on from TRI, uh RAI, uh, you know, it's just Mark Rabert from Basa Dynamics.
• 8:13
  SPEAKER_00: Do you feel that places like that are or at least are attempting to sort of fill a similar role?
• 8:20
  SPEAKER_01: I think uh yeah, there are places that definitely attempt to do that, but the reality of the world of technology, as you know, Brian, is that it's extremely competitive now.
• 8:30
  SPEAKER_01: And competitive in the sense that you have to be extremely efficient, you have to use your budget accordingly, and you need to make sure you stay on top of the competition, especially at global level.
• 8:41
  SPEAKER_01: So that uh drives a lot of even giant companies to focus on next quarter, this year, and beyond.
• 8:49
  SPEAKER_01: And that is how how how the market drives a lot of behaviors.
• 8:54
  SPEAKER_01: So a lot of these institutes, of course, are trying to carve out some areas that you wanted to look beyond tomorrow.
• 9:01
  SPEAKER_01: But uh I think uh in this environment and this competitive world of technology, uh it's becoming more challenging to do everything under one roof.
• 9:11
  SPEAKER_01: So you need to tie it with universities, but at the same time, you have the understanding of what it takes to bring it to a company.
• 9:17
  SPEAKER_01: So you cannot build a Bell Lab like in any companies today with the critical mass of thousands of scientists.
• 9:24
  SPEAKER_01: I mean, that is not affordable in a highly competitive world unless you have a monopoly on something.
• 9:30
  SPEAKER_01: That's what how it had used to be like there back then.
• 9:33
  SPEAKER_01: So I think uh the exact model is not practical in this environment, but we can learn from that and use resources inside the company and outside and making sure that you tie these pieces together.
• 9:48
  SPEAKER_01: So today the art of making a virtual bear lab is very important, meaning that how to connect the dots inside and outside to leverage all of that.
• 9:56
  SPEAKER_01: Because, as I said, affording such a big critical mass of top scientists is very difficult to replicate today.
• 10:04
  SPEAKER_01: And the diversity of topics was amazing.
• 10:06
  SPEAKER_01: It was from software.
• 10:07
  SPEAKER_01: I mean, can you can imagine C language was invented there and laser and then communication and transistors can see from device to software.
• 10:16
  SPEAKER_01: Who can do that in this world with these competitive environments?
• 10:19
  SPEAKER_01: That's that's tough.
• 10:20
  SPEAKER_01: But uh, we can definitely come up with more creative ideas to to replicate as much as possible in a different way.
• 10:29
  SPEAKER_00: Yeah, that that that's an interesting point.
• 10:31
  SPEAKER_00: Um, you know, and you were talking about uh theoretical physics, for example, and obviously for several years now we've been talking about um quantum computing, but these are there are these um, you know, at least if you look at present technology, seemingly unrelated fields and unrelated sciences, and you have to really be forward-looking to start pulling those people in.
• 10:54
  SPEAKER_00: So, like you said, I mean, that's that's what takes tremendous resources is pulling in the people, the brains, and the resources that might not pay off until like 10, 15 years down the road.
• 11:06
  SPEAKER_01: Exactly.
• 11:07
  SPEAKER_01: And and on top of that, uh even 10 years, 15 years, I mean, time access is one of the dimensions, but also you'll be surprised in your journey that how many competing technologies may show up.
• 11:20
  SPEAKER_01: And you need to react to that.
• 11:22
  SPEAKER_01: I mean, uh uh in the past, I mean, if you look at the invention of transistors, many top physicists in Bell Labs spent ages to work on it.
• 11:31
  SPEAKER_01: They had the strong critical mass to work on it.
• 11:34
  SPEAKER_01: But in today's world, there are competing groups and technologies and people, talented people around the world.
• 11:41
  SPEAKER_01: And you will never, I mean, you never know in five years who's gonna come up with an idea that makes your idea less relevant or the other.
• 11:48
  SPEAKER_01: So that's kind of it's a very different environment.
• 11:51
  SPEAKER_01: We need to be much more reactive to what's happening around out there.
• 11:55
  SPEAKER_01: Don't get too um siloed in our thinking so that, okay, what I'm working on is gonna make the world different, and then ignoring competition.
• 12:04
  SPEAKER_01: Because uh the innovation, I mean, back then Bell Labs was the only place with that critical mass.
• 12:10
  SPEAKER_01: Now we have places that are almost kind of leveled playing field.
• 12:14
  SPEAKER_01: So you need to also be somehow paranoid that, you know, who else is gonna come and offer something that might be more compelling.
• 12:24
  SPEAKER_00: This is, you know, and I'm I'm saying this is very much an outsider, but I've been watching um, you know, uh I've been writing about technology for a long time, largely consumer, and obviously now a lot about AI and robotics.
• 12:38
  SPEAKER_00: But I I think what you're talking about is particularly relevant when we're discussing chips and silicon.
• 12:46
  SPEAKER_00: And I think that your career is a really interesting example of this because I was going back and I was doing, you know, I mean, it's this is something, this is something interesting, you know, especially in kind of the corporate world and and in the tech technology world.
• 13:01
  SPEAKER_00: You go back, you look at somebody's resume, that doesn't really tell the story.
• 13:05
  SPEAKER_00: You have to go back, you have to look at the mergers and the acquisitions, and that starts to tell the story.
• 13:12
  SPEAKER_00: Out of Bell Labs, you start a company.
• 13:15
  SPEAKER_00: That company, uh I think it was primarily like mobile communication, right?
• 13:20
  SPEAKER_00: What you were working on.
• 13:22
  SPEAKER_01: Wi-Fi, linguistics, signal.
• 13:23
  SPEAKER_01: Yeah, yeah.
• 13:24
  SPEAKER_00: Yeah, what you were working on in a sense around Bell, you get acquired by a company that gets acquired by TI, but uh I want to back up.
• 13:34
  SPEAKER_00: So that company was national semiconductor, and they were at the at the time, their big competition or who they were attempting to compete against was Intel.
• 13:44
  SPEAKER_01: Absolutely.
• 13:45
  SPEAKER_01: I mean, National Semi and Intel were two pillars of uh uh chip development in the in the in the Silicon Valley in the world.
• 13:52
  SPEAKER_01: I mean, Intel was digital and national was analog, and they were all they almost started at the same time in the late 50s and uh early 60s.
• 14:00
  SPEAKER_01: So these two giants, Intel and National, working on two segments of market, and there were some overlaps over time.
• 14:08
  SPEAKER_01: And interestingly, when National Semi, after two and a half years of incubation of my startup, approached us, I was thinking, yeah, okay, I moved from a giant company like Bell Labs to a startup and now going back to another big company.
• 14:22
  SPEAKER_01: That would be an interesting journey.
• 14:24
  SPEAKER_01: And uh I spent quite a bit of time in National Semiconductor, I always had some side engagement with UC Berkeley teaching at UC Berkeley later at Stanford, so always kept my academic link uh intact because I felt I always believe that in that engagement is very helpful to make sure that you are not only motivated by all the talents and excitements of research, but also understand the latest and greatest that people have been, I mean, topics that people are working on, and uh use that as a point of inspiration for what you can do in the company.
• 14:59
  SPEAKER_01: So, yeah, National Semiconductor was the iconic company in Silicon Valley in analog and mixed signal and many other areas.
• 15:08
  SPEAKER_01: And interestingly, was a competition for TI and Texas Instrument acquired National Semiconductor.
• 15:13
  SPEAKER_01: So big brother took over the younger brother, and it became like the number one, I mean, now the largest analog mixed signal company in the world.
• 15:23
  SPEAKER_00: Yeah, and and I bring this up again because uh here's here's one of the many reasons why all this is is is interesting when you take the thousand-foot view of it all, is because again, sort of like going back and reading um about what was going on with the companies, it seems like around that time um uh National Semiconductor was also sort of stepping away from smartphones, which was which had been a big thing for them.
• 15:49
  SPEAKER_00: And then obviously, you know, there was at that time, like 10, 15 years ago, when NVIDIA was stepping away from smartphones.
• 15:57
  SPEAKER_00: They started making an investment in AI and robotics.
• 16:01
  SPEAKER_00: That you know, that was a really long shot, far-off investment.
• 16:05
  SPEAKER_00: Um, Intel lost the mobile battle.
• 16:08
  SPEAKER_00: And and it's really this just it's this this ebbing and flowing and and and up and down.
• 16:13
  SPEAKER_00: And it's it I it I don't know.
• 16:15
  SPEAKER_00: I don't know if there's another field that's quite like the chip field.
• 16:20
  SPEAKER_01: Yeah, it's uh chip field has one of the most interesting characteristics that is unique, unlike software, that we can talk about later.
• 16:28
  SPEAKER_01: But uh I believe that there is no other technology in human history as creative and as innovative as semiconductor.
• 16:38
  SPEAKER_01: Because if you think about it, over the last many decades we managed to get these chips a million times more efficient, but at almost the same cost in terms of the per the real estate of the silicon.
• 16:50
  SPEAKER_01: And that is by amazing combination of innovations from material science to manufacturing to kind of like tool manufacturing to lithography.
• 17:02
  SPEAKER_01: So there is so many disciplines pushing the envelope of the science.
• 17:06
  SPEAKER_01: I mean, if you look at EUVs or technologies that are used there, these are massive amounts of technology innovations.
• 17:12
  SPEAKER_01: How to make these billions of transistors on a tiny chip?
• 17:16
  SPEAKER_01: I mean, these are amazing accomplishments that really, because we managed to scale, we managed to make the wafers bigger, we managed to get the throughput almost uh you know perfectly done in a way that we can get highest performance and highest cost and highest value out of these chips.
• 17:35
  SPEAKER_01: So combination of all these innovations fueling everything around us.
• 17:40
  SPEAKER_01: If you look at AI, I mean, because of the advances in chips that now you can afford, because AI concept is not new, it was started in the 1950s at MIT.
• 17:49
  SPEAKER_01: I mean, the first AI lab was in 1950-something at MIT.
• 17:52
  SPEAKER_01: So the concept has been there, but the amount of processing and of course the data and all the algorithms, I mean, all are based on this foundation of uh this semiconductor technology.
• 18:05
  SPEAKER_01: I mean, if you look at the autonomy and robotics, if you look at all the gadgets and devices around you, all the intelligence, because these bits and waveforms need to be processed on a physical environment.
• 18:16
  SPEAKER_01: It doesn't come from like uh air, right?
• 18:19
  SPEAKER_01: So I think semiconductor has fueled almost every innovation.
• 18:23
  SPEAKER_01: I was talking to a major car company, and they said that, and this is a company's been around forever, and they were saying that 90 more than 90% of their innovation is on the electronics of the car, not on the mechanics.
• 18:35
  SPEAKER_01: So you can see that semiconductor is fueling uh the innovation across the board, including some of the most exciting stuff like AI and quantum computing.
• 18:46
  SPEAKER_00: Well, that I mean that's a the you know, I and I think a really good um a really good illustration of that is when you know during the pandemic, there were these really like clear visuals of flyovers in the Detroit area, and all of these racetracks were covered with all of these cars, and the cars were basically finished, but for one thing, they couldn't get the chips over, right?
• 19:11
  SPEAKER_00: And the chips, the chips are just so essential now to every single part of the car that you just you can't have a car in 2026 without the chips.
• 19:22
  SPEAKER_01: Absolutely.
• 19:22
  SPEAKER_01: We called it golden screw back then.
• 19:24
  SPEAKER_01: I mean, you know, you you need to have this uh tiny chip to get this whole massive car or truck to get to the finish line.
• 19:31
  SPEAKER_01: And the number of chips in the car and intelligence in the car, you can you can see how much of autonomy from ADAS, I mean the driver assistance, or from safety angle, from communication angle, the number of chips are amazing and number of radar chips that TI is offering in automotive inside the car and outside the car, just you know, growing.
• 19:52
  SPEAKER_01: And uh again, it all goes to the fact that semiconductor industry thrives on a scale and can man by I mean right everybody in the in the in the field, like two years ago, uh giving forecast that you know by 2030 this market would be more than a trillion dollars.
• 20:10
  SPEAKER_01: Guess what happened?
• 20:11
  SPEAKER_01: This year is gonna be more than a trillion dollars.
• 20:13
  SPEAKER_01: So you're four years early in our growth because of the fact that more data centers, more uh autonomy, more connected devices are are growing.
• 20:24
  SPEAKER_01: And in TI, because of that kind of trend, we invested a lot on manufacturing because we believe that having more manufacturing capability, advanced fabs, give us a chance to take advantage of these opportunities in the fast growth markets in semiconductor.
• 20:41
  SPEAKER_00: Yeah, I mean, I don't I don't want to linger around the pandemic for too long, but you know, now that I'm thinking about these again, these like very vivid visuals of all the cars out there and all the conversations that we were having around supply chain, um, around manufacturing.
• 20:56
  SPEAKER_00: Um, you know, obviously there were a lot of big political conversations around uh localizing manufacturing, decentralizing manufacturing, bringing more chip manufacturing to the US.
• 21:07
  SPEAKER_00: And I and I'm curious as you as you look at a company like TI, I mean, ultimately, how profound of an impact did those few years and did those supply chain shortages have over the way you do business?
• 21:23
  SPEAKER_01: Absolutely.
• 21:23
  SPEAKER_01: I mean, the Uranian pandemic, I remember uh uh we felt that you know manufacturing and supply chain in our business is of ultimate value and importance.
• 21:35
  SPEAKER_01: So, right coming out of the pandemic, we expanded our manufacturing, putting billions of dollars, and mainly in US, I mean, almost all of them actually in US, and uh creating this huge manufacturing capacity.
• 21:49
  SPEAKER_01: And sometimes we are questioned that look, because as you know, it's I mean, semiconductor markets has this cyclic behavior.
• 21:55
  SPEAKER_01: I mean, always, I mean, every four four years we have uh peak and throughout.
• 22:00
  SPEAKER_01: And interesting things that uh building those fabs is a four or five years' experience with billions of dollars.
• 22:06
  SPEAKER_01: So you have to start at the trough so that we can catch the next pick.
• 22:10
  SPEAKER_01: But then you get maximum pressure that why are we spending money when the market is down because you expect the market to go up.
• 22:17
  SPEAKER_01: And you can see that uh vision and that uh uh understanding of you know the demand for semiconductor keeps growing and even at a faster pace um is paying off.
• 22:31
  SPEAKER_01: And we believe that even it's going to be more urgent to have this secure supply chain as well as the manufacturing capacity and technology, a competitive technology, to respond to the demands in the market.
• 22:47
  SPEAKER_00: Yeah, I mean, talking of predictions that happened maybe a little bit, maybe not earlier, but came on more quickly than expected.
• 22:55
  SPEAKER_00: I you know, the I mean, obviously, people like you and I, we've been talking about AI for a long time.
• 23:00
  SPEAKER_00: I've been talking and writing about robotics for a long time, but it really did seem to enter the mainstream conversation almost overnight, you know, and and a lot of that was um uh obviously LLM's um uh generative AI and uh just the ability of the average person to sort of like get on a computer and and interact with it and and create something.
• 23:21
  SPEAKER_00: Um obviously, kind of ostensibly, you know, a big part of the reason why we're having this conversation right now is because of the the plays that TI has had in and around robotics and and and humanoids.
• 23:33
  SPEAKER_00: Uh and I'm curious, as we sort of like look into the past a little bit, you know, I mentioned NVIDIA making that pivot from mobile to robotics.
• 23:43
  SPEAKER_00: Um in the lead up until now, in in the past 10 to 15 years, you know, what what if anything has the company been doing to prepare itself for this moment?
• 23:55
  SPEAKER_01: Yeah, actually it's interesting because when you look at these uh you know highly uh you know uh strong and uh popular trends like AI or IoT or robotics, uh I believe that there are fundamentals that they keep getting better and better.
• 24:16
  SPEAKER_01: And at some point they are good enough to create these opportunities.
• 24:19
  SPEAKER_01: Let me give an example.
• 24:21
  SPEAKER_01: When we design these big chips, being a GPU or any big processors, sometimes you are limited by how much power you can deliver to these chips, not just by the number of transistors.
• 24:32
  SPEAKER_01: So that process, you call it AI or you call it graphic processor for videos.
• 24:37
  SPEAKER_01: I mean, power delivery and efficient power delivery is a trend that gets getting more and more critical for all these big chips.
• 24:45
  SPEAKER_01: Or if you look at connectivity and bandwidth, you need to have that bandwidth because you're connected.
• 24:50
  SPEAKER_01: I mean, AI thrives on data, so you need to connect things.
• 24:54
  SPEAKER_01: Uh so um at some point, copper is gonna run out of bandwidth and you need to go to photonics.
• 25:01
  SPEAKER_01: So these trends is kind of, we cannot tell when exactly it happens, but can see the trend very clearly.
• 25:07
  SPEAKER_01: You need more power density, you need more bandwidth per joules, you need more accuracy in your timing and clock because these devices are communicating and switching at faster rates.
• 25:18
  SPEAKER_01: So these trends are there and have been there for many decades.
• 25:22
  SPEAKER_01: Sometimes they accelerate.
• 25:24
  SPEAKER_01: So one of the things that in TI and in Kilby Labs you always have been mindful of, nobody can predict the next market in 10 years.
• 25:34
  SPEAKER_01: I mean, you can always have guesses.
• 25:36
  SPEAKER_01: But you what you can see is that the trends are very clearly there.
• 25:39
  SPEAKER_01: For example, offering higher power density, offering higher bandwidth, offering higher accuracy on a clock.
• 25:47
  SPEAKER_01: These are very important.
• 25:48
  SPEAKER_01: And guess what happens?
• 25:49
  SPEAKER_01: Now that we have these massive data centers, one of the bottlenecks is how to deliver power to these giant chips.
• 25:56
  SPEAKER_01: One of the challenges is how these chips can communicate on a super high speed of the uh uh terabits per second and so on.
• 26:04
  SPEAKER_01: So I think uh for us, I mean general, we cannot we cannot just predict the future, especially 10 years from now, but we can tell the trends are there and we invest such that we can react to those demands as it comes.
• 26:22
  SPEAKER_00: Yeah, it's funny.
• 26:23
  SPEAKER_00: I was going back and and and watching some some talks you had given over the years, and you gave one in 2018 where you were talking about the sort of massive demands of these data centers, and I was just thinking, oh poor boy, you know, just you just you wait.
• 26:39
  SPEAKER_00: You know, obviously it this is like this is a huge conversation right now, right?
• 26:42
  SPEAKER_00: And this is this this is the big thing about AI.
• 26:44
  SPEAKER_00: All these they're being um servers are being built everywhere, uh, huge huge, huge power demands.
• 26:52
  SPEAKER_00: I I read something the other day.
• 26:54
  SPEAKER_00: I don't want to quote any numbers because I don't want to get it wrong, but you know, that it's gonna be hard, obviously it's gonna be harder for a lot of these companies to hit the climate pledges that they've had over the last few years, because it's just like Microsoft was the one that that that came up.
• 27:09
  SPEAKER_00: Um but I will say that one of the things that jumped out at me when watching um a video around the opening of Kilby Labs was it seemed that the big focus on that early research was low power.
• 27:27
  SPEAKER_00: So, you know, it it is is that something that that that that you can square?
• 27:33
  SPEAKER_00: Is there a way to move forward, to have more and more and more compute demands, to see that this is gonna be more and more research intensive, to see that at some point like copper is just not gonna do the trick anymore and you know just not get more doom and gloom about it.
• 27:52
  SPEAKER_01: Absolutely, absolutely.
• 27:53
  SPEAKER_01: Again, as I mentioned, I mean, this power density, it shows itself in your mobile.
• 28:00
  SPEAKER_01: I mean, you want it to you know, want it to have a lot more functionality.
• 28:04
  SPEAKER_01: As you saw, there is an announcement that now you have multiple GPUs in a laptop, or in uh eventually you can have it in your uh mobile phone.
• 28:14
  SPEAKER_01: So that means demand for power from portable devices all the way to data center is insatiable.
• 28:20
  SPEAKER_01: I mean, it keeps going up.
• 28:21
  SPEAKER_01: You have a very challenging environment to deliver as much power to run these devices.
• 28:26
  SPEAKER_01: I mean, these are like a few kilowatt devices that now we are talking about, a single chip.
• 28:31
  SPEAKER_01: So uh from mobile devices, and it could be even uh your tiny um smartwatch that needs more power to do more processing on site uh in the future.
• 28:43
  SPEAKER_01: I think uh even portable devices have some level of AI engine to do this edge processing or HII in addition to data centers.
• 28:53
  SPEAKER_01: So demand for power density just is keeps rising exponentially.
• 28:58
  SPEAKER_01: And I had a talk on that like about 10 years ago that I called it convergence of exponentials.
• 29:04
  SPEAKER_01: I mean, how the number of gates and amount of power and bandwidth and wireless connectivity all are exponentially going up, but with a different uh rate.
• 29:15
  SPEAKER_01: And the question is how to fill those gaps.
• 29:18
  SPEAKER_01: And that is one of the interesting things that we see.
• 29:22
  SPEAKER_01: The pace of this data center's power demand is moving so I mean so fast and so amazingly dominant that we really need to spend a lot of time on even offering more uh more efficient power delivery systems.
• 29:37
  SPEAKER_01: And we are heavily involved in that, and we are heavily involved with many of these chip manufacturers that are designing chips for big data centers and also for data center companies.
• 29:48
  SPEAKER_01: So I would say these hyperscalers, to me, one of the exciting things that it in the in in this day and time is that hyperscalers and data centers are pushing technologies limits to the edge.
• 30:03
  SPEAKER_01: And we need to come up with new tricks to respond to those demands across the board.
• 30:09
  SPEAKER_01: From uh these integrations, 3D integrations, beyond CMOS, new devices for power management, moving from copper to fiber, and even on the board, or eventually maybe some at some point on the chip, and uh offering highest accuracy of timing and clock references.
• 30:29
  SPEAKER_01: So pushing everything to the limit, and then the rest of the market can take advantage of that.
• 30:34
  SPEAKER_01: Because once you prove that concept in a data center, you can borrow it for automotive and others.
• 30:40
  SPEAKER_01: So data centers becoming one of the leading drivers of innovation across the board in different disciplines of the semiconductor.
• 30:49
  SPEAKER_00: Yeah, I wonder too.
• 30:51
  SPEAKER_00: Um I mean, I'm I'm hopeful when I look at a lot of the research that's happening around transformers and AI, that uh up till now we um researchers have taken a really uh spaghetti against the wall approach to AI with these massive, massive models.
• 31:06
  SPEAKER_00: And in a lot of cases, we're finding you know some accuracy, a lot of accuracy can be found.
• 31:13
  SPEAKER_00: And obviously, we're finding um uh we're getting a little more insight into the black box, into how these things actually work.
• 31:19
  SPEAKER_00: And and hopefully, my hope is that that signals a future where that won't require as much constant compute power and as you know, just as much power period as we currently require.
• 31:37
  SPEAKER_01: No, I I fully agree with you.
• 31:39
  SPEAKER_01: I think for the next five to ten years, one of the most exciting innovation areas are the algorithms that are much more efficient in terms of resource requirements.
• 31:49
  SPEAKER_01: I mean, we got our surprise last year by these reasoning models that are much more uh resource efficient and can give high accuracy.
• 31:58
  SPEAKER_01: And there has been a lot of publications and work on uh different models.
• 32:02
  SPEAKER_01: So, and people are talking about small language models versus large language models.
• 32:07
  SPEAKER_01: I mean, how many parameters beyond which you get so much efficiency or less efficiency?
• 32:11
  SPEAKER_01: So I think the algorithms to make it more efficient and less resource-demanding, I think would be one of the most exciting areas.
• 32:20
  SPEAKER_01: And there's a lot of room for that.
• 32:22
  SPEAKER_01: I don't know if you have kept up with some of the latest, but uh, some of these LRM versus LLM versus SLM showing some really promising results for innovation in the algorithms.
• 32:34
  SPEAKER_01: I mean, it's like kind of like our brain, the way that we are doing inference.
• 32:38
  SPEAKER_01: It's not, I mean, language is different from reasoning from other tasks that we are doing, and each part of the brain is optimized for that kind of inference.
• 32:46
  SPEAKER_01: So we can get inspired by that to offer much more optimized algorithms.
• 32:52
  SPEAKER_01: Right now, literally, we are putting massive amount of processing for these LLMs, which is really interesting.
• 32:59
  SPEAKER_01: But down the road, I can see uh algorithm innovations in algorithm also can help us to move to the next level of efficiency.
• 33:07
  SPEAKER_00: I I wanted to get back to this uh I idea of anticipating technologies.
• 33:12
  SPEAKER_00: And I think to kind of put a finer point on it, effectively what you were saying is that um as a company, what TI is really anticipating is less the end product for the consumer and more the needs of the technology.
• 33:29
  SPEAKER_00: Is that a fair assessment?
• 33:31
  SPEAKER_01: Uh I would say TI uh is very unique in the sense that the diversity of technologies and diversity of markets we are addressing is quite amazing.
• 33:41
  SPEAKER_01: We are looking at everything from consumer to data centers to medical to industrial automotive, everything that we are looking at.
• 33:49
  SPEAKER_01: And that diversity of industrial automotive data center and others gives us resilience, but also technologies, some of the foundational technologies can be leveraged across different markets.
• 34:01
  SPEAKER_01: For example, if you have a high voltage technology or devices that are dealing with high voltage power delivery, you can use it in a data center, but also the EV market can benefit from that as well.
• 34:13
  SPEAKER_01: So the beauty of having this kind of cross-pollination of the technologies give us more resilience in terms of end applications.
• 34:21
  SPEAKER_01: But at the end of the day, each of these markets have their own unique set of requirements as well that you need to meet.
• 34:27
  SPEAKER_01: But the foundational technologies and the products we have can serve many of these markets.
• 34:34
  SPEAKER_01: And that is one of the signs of release.
• 34:36
  SPEAKER_01: I mean, that's one of the unique capabilities of a company like TI.
• 34:41
  SPEAKER_00: Yeah, I mean, this is a conversation we have all the time.
• 34:43
  SPEAKER_00: And the classic example of this is the space race example, right?
• 34:47
  SPEAKER_00: Like all the money that goes into getting to the moon, and then we that's why we have Velcro.
• 34:51
  SPEAKER_00: Like that, you know, like you know, and that and that's you know, the kind of the closest, maybe that maybe that's like, I don't know, a model for like of like a Bell Labs or something, like, you know, just research, kind of research for and science for for science sake.
• 35:06
  SPEAKER_00: Um, but we've seen some to an extent, we've seen something similar happen with uh autonomous driving, right?
• 35:14
  SPEAKER_00: I mean, it it's taken a lot longer to get there.
• 35:16
  SPEAKER_00: We're finally getting there, but we've also I keep saying we're not having this conversation about humanoid robots, but but also just like uh general or multi-purpose robots if self-driving cars aren't there, if the hardware and the software hadn't hadn't been developed for them.
• 35:34
  SPEAKER_00: And I and I think one of the things that you were alluding to is um maybe to an even larger degree than most of us are even talking about at this point, that that sort of innovation is really happening on the data center side.
• 35:49
  SPEAKER_01: Absolutely.
• 35:50
  SPEAKER_01: I mean, again, robotics, you refer to that, is another example.
• 35:53
  SPEAKER_01: This concept has been around forever.
• 35:55
  SPEAKER_01: I mean, for many decades, people have been talking about uh robots, and I know that uh many companies started this robotic development, but uh the level of technology maturity you need to have a robot in the mainstream, meaning that motor drive technology, sensing technology, analytics technology.
• 36:15
  SPEAKER_01: Reliability, repeatability, I mean, all of these need to converge together.
• 36:20
  SPEAKER_01: And sometimes in industry, we uh we think that these things can happen overnight.
• 36:25
  SPEAKER_01: But if you think about it, I mean, look at smartphone.
• 36:28
  SPEAKER_01: I mean, the pieces of technology that created a smartphone were built over many decades.
• 36:33
  SPEAKER_01: I mean, display got so much better that you can have a touch display, and then connectivity became so good.
• 36:39
  SPEAKER_01: And there were many attempts to do a smartphone before these technologies were mature enough, and it failed.
• 36:44
  SPEAKER_00: People forget how bad touch displays were before the guy.
• 36:47
  SPEAKER_01: Exactly, exactly.
• 36:48
  SPEAKER_01: Yeah, but take it for granted that how much of these different vectors worked independently and then converged.
• 36:56
  SPEAKER_01: I mean, the even glass technology that is using the smartphone, I mean, that went through its own.
• 37:01
  SPEAKER_01: So I think uh these technologies have many components that almost are growing together in different contexts and then converge at some point.
• 37:12
  SPEAKER_01: And uh, of course, uh you see at that point, I mean, uh you see a huge difference.
• 37:17
  SPEAKER_01: So, for example, autonomous driving.
• 37:20
  SPEAKER_01: I mean, I remember I was sitting in a panel 12 years ago, 10 years ago, with some uh key car manufacturers and companies, and everybody was saying, Oh, within a couple of years we are gonna have autonomous driving.
• 37:32
  SPEAKER_01: And now, 10 years past that point, how many cars you see in the street that are fully autonomous?
• 37:38
  SPEAKER_01: I mean, San Francisco you see way more and others.
• 37:40
  SPEAKER_01: I mean, but uh still it's not made, it's getting there, but it takes a lot longer because many things can converge to become uh become uh kind of like a magic touch of okay, smartphone data centers.
• 37:55
  SPEAKER_01: I mean, data centers you can imagine offering kilowatts of power to a chip.
• 38:01
  SPEAKER_01: That was just unimaginable, like 10 years ago.
• 38:04
  SPEAKER_01: I mean, how much energy efficiency you need to provide, and how much of the innovation in your power delivery?
• 38:10
  SPEAKER_01: Uh, how to integrate memory and processors, how to communicate across processors and racks and the blades.
• 38:18
  SPEAKER_01: I mean, these are things that uh really are built on many decades of innovation in that field and then converge together to make the magic happen.
• 38:28
  SPEAKER_00: I think about this a lot in terms of the the there's a reason why it's so hard to predict the weather.
• 38:34
  SPEAKER_00: Because there are, you know, there are so many different factors that go into the climate that it's you know in impossible, certainly impossible for one person to uh to work through them all.
• 38:47
  SPEAKER_00: There's, you know, there's no like one equation you could sit down to work through them.
• 38:50
  SPEAKER_00: And I think it's very similar with technologies um wherein you know, some somebody focuses on a certain set of really hard problems, and that's why they think there's going to be a certain timeline.
• 39:02
  SPEAKER_00: But there are just so many things you don't account for.
• 39:05
  SPEAKER_00: I mean, you know, obviously in in self-driving cars, you know, um regulation has been a big part of it.
• 39:10
  SPEAKER_00: But then there's all these like long tail problems.
• 39:13
  SPEAKER_00: There are um, you know, unexpected issues that that have popped up.
• 39:18
  SPEAKER_00: And it's just it isn't possible.
• 39:20
  SPEAKER_00: You know, that I the smart people I think can see the writing on the wall as far as this thing is coming eventually, but like don't don't put money down on the coming in a certain set of years.
• 39:29
  SPEAKER_00: And I've talked to consistently talked to incredibly smart people and uh have asked them questions.
• 39:35
  SPEAKER_00: I was asking the AGI question, you know, 10 years ago, and it just kept it was five to ten years every year.
• 39:42
  SPEAKER_00: Five to ten years, five to ten years, just kept getting pushed back.
• 39:46
  SPEAKER_01: Yeah.
• 39:46
  SPEAKER_01: I mean, and in addition to all that, uh, what is interesting, and I've experienced that in my professional life, is that sometimes uh scientists say that okay, here's the limit.
• 39:57
  SPEAKER_01: You cannot go beyond that limit because it's built on some physics law.
• 40:01
  SPEAKER_01: And it's amazing how often I've seen engineers surprise scientists to find a way to go around that.
• 40:06
  SPEAKER_01: Okay, here is the Boltzmann constant.
• 40:08
  SPEAKER_01: You cannot just beat that, but then it's even other way they can do things.
• 40:12
  SPEAKER_01: So I think the idea that a lot of smart people working on technologies and come up with the ways that surprise the scientists.
• 40:21
  SPEAKER_01: I remember um maybe 10, 15 years ago, there was a very famous physicist, okay, you cannot go below 55 nanometers.
• 40:29
  SPEAKER_01: That's gonna violate some basic uh fundamentals of the physics law.
• 40:34
  SPEAKER_01: And guess what?
• 40:35
  SPEAKER_01: We are now at two nanometers today because we found a way to go around that.
• 40:38
  SPEAKER_01: I mean, so there are amazing um pipeline of innovative ideas, especially on AI, which is highly multidisciplinary.
• 40:46
  SPEAKER_01: Again, the three pillars of AI is, of course, massive processing of data centers, but also massive amount of data, especially physical AI that comes to the picture.
• 40:56
  SPEAKER_01: And the third one is the algorithms that are really processing that data.
• 40:59
  SPEAKER_01: And all three of them are going forward in a fast pace.
• 41:03
  SPEAKER_01: And it's gonna change uh, I mean, it's it's gonna be a very different picture 10 years from now, uh, based on what I see as far as the trends and opportunities and applications.
• 41:14
  SPEAKER_01: But uh, I don't think any of us can paint exactly what 10 years from now is gonna look like.
• 41:19
  SPEAKER_00: Yeah, I really I one of the things I've started really loving about this job, I was a humanities major, you know, I was a creative writing major.
• 41:27
  SPEAKER_00: Um, you know, technology is something that I've I've I've come into uh from a uh uh a journalistic angle.
• 41:35
  SPEAKER_00: And especially as I started covering AI and robotics, one of the things that I love writing about and talking to people about is the surprise, you know, and and it's very interesting to hear very technical and very scientific-minded, very smart people talk about things like magic.
• 41:54
  SPEAKER_00: Because, you know, it's that I I know this gets thrown around a lot with the the Arthur C.
• 41:59
  SPEAKER_00: Clark quote about, you know, it was a sufficiently uh uh uh evolved technology is indistinguishable from magic.
• 42:06
  SPEAKER_00: That and and and I think that's part of what makes specifically has has made um the advances in AI and LLM so incredible, is that you've got the people who are working on them that are completely surprised that they work the way they do.
• 42:21
  SPEAKER_01: That's right.
• 42:22
  SPEAKER_01: That's right.
• 42:22
  SPEAKER_01: I mean, I I if you look at uh let's say AI by uh evolution, I mean reality AI, as we talked before, it was not a new thing, it's been evolving for many decades.
• 42:33
  SPEAKER_01: But when this generative AI came to the picture and the concept of uh this LLM and topics that are kind of adding a different spin to the traditional AI algorithm, it all of a sudden changed everything.
• 42:46
  SPEAKER_01: And now you can see that when you have this massive number of parameters or tokens in your algorithms, you need a massive amount of processing to deal with it, and you need a massive amount of processing.
• 42:58
  SPEAKER_01: So, how do you uh how do you offer this memory storage to deal with this kind of processing demand?
• 43:05
  SPEAKER_01: And then how do you power them up?
• 43:07
  SPEAKER_01: So it kind of ripples through and drives everything to the limit, and then uh a lot of smart people come up with new ideas.
• 43:15
  SPEAKER_01: So we are trying at TI to catch all these vectors and making sure that uh the foundation is there and the new devices, new technologies, new circuits, and the manufacturing to support that is there to respond to these opportunities and get take advantage of them.
• 43:36
  SPEAKER_00: At what point was it clear that uh you know, okay, we we really need to get serious and you know, and and talk about these partnerships uh in and around robotics and you know, in and around things like humanoids?
• 43:50
  SPEAKER_01: Again, it's interesting, uh Brian.
• 43:53
  SPEAKER_01: I I I think that trend we could see it, I mean, for many years and decades.
• 43:59
  SPEAKER_01: It was not like At one point, I'd say okay, we realize it's big, it's just an evolution more than a revolution in a lot of these technologies.
• 44:07
  SPEAKER_01: Because I remember we started a project on high voltage devices, because these new devices have different material, and they're very challenging to make sure that not only they can perform, but also they are manufacturable.
• 44:20
  SPEAKER_01: So we spent a lot of time and money on that, and our first goal was the smart grid.
• 44:26
  SPEAKER_01: I mean, that was our goal when we started that.
• 44:29
  SPEAKER_01: And then very soon a new market came up for that.
• 44:33
  SPEAKER_01: That is electric vehicle, because they need high voltage.
• 44:36
  SPEAKER_01: I mean 800 or 600 volt and 800 volt.
• 44:40
  SPEAKER_01: And then after that, data centers came.
• 44:42
  SPEAKER_01: So, you know, so we started this technology development for a very different purpose.
• 44:47
  SPEAKER_01: And we saw that these new technologies and trends are coming.
• 44:50
  SPEAKER_01: And we said, okay, now how do we respond to these by using this technology?
• 44:55
  SPEAKER_01: So I think the trends are there, but the key point is that you need to have a 360-degree view of what is pushing the next wave and what is driving the next opportunity.
• 45:09
  SPEAKER_01: And uh, I always say in our research environment, and being in RD and research since Bell Labs time, the worst thing can happen for a researcher is to get so to fall so much in love with their own ideas, they ignore the reality of what's going on around them.
• 45:23
  SPEAKER_01: And say, oh, I've worked on this for five years, I have to make it work.
• 45:27
  SPEAKER_01: But what if you need to change the direction five plus minus five degrees, or maybe there is a competing technology that you need to pay attention to?
• 45:35
  SPEAKER_01: So I think the challenge is uh really to catch these opportunities, but you need to have the foundations there.
• 45:43
  SPEAKER_01: I mean, just give an example.
• 45:45
  SPEAKER_01: Even when you wake up to the fact that AI is going to be big, a new device takes five, six years to develop.
• 45:51
  SPEAKER_01: It's not like tomorrow.
• 45:52
  SPEAKER_01: So if you just wait for the next market to start something, then you're always behind the curve and not having like opportunity to capture these uh growth opportunities.
• 46:04
  SPEAKER_01: So that's why I think we need to look at these trends and these exponentials from a foundational technology viewpoint and then react to the markets that they come, they become more important uh rather than uh having just uh uh a kind of like a uh impulsive reaction to what the market is doing.
• 46:25
  SPEAKER_00: Yeah, and we're we're getting close to time, so we we can close on this.
• 46:28
  SPEAKER_00: But I, you know, you were talking about well, we were we were talking about uh uh reliability and repeatability, and um the point you get to, you know, a technology being iterative and the point that you get to when you can really believe in this technology at a mass scale, you know, for for uh for industry, for manufacturing, that means something that is like highly repeatable.
• 46:54
  SPEAKER_00: But for humanoids, that means something that's highly repeatable, but also could do a lot of different things.
• 46:58
  SPEAKER_00: That's big part of the reason why this is so difficult.
• 47:01
  SPEAKER_00: Um, and and that, I feel like that's gonna be the time when when a company like TI is like, all right, we're there, like let's start looking at getting these things, you know, we're we're looking at decentralizing manufacturing, uh, we're looking at getting more centers, let's let's start taking a look at these humanoids, like that's at a point where you know things have gotten reliable.
• 47:21
  SPEAKER_01: Yeah, absolutely.
• 47:22
  SPEAKER_01: So let's let's take robotic and manufacturing.
• 47:25
  SPEAKER_01: It's been around for a long while, but it was always in a cage and doing repetitive work, not not reacting to the environment.
• 47:33
  SPEAKER_01: Pretty much take the some from point A to B, and you put it in a cage because they had no awareness of surrounding, it has to be safe.
• 47:41
  SPEAKER_00: Now they're heavier metal.
• 47:43
  SPEAKER_01: Yeah, yeah, heavy metal.
• 47:44
  SPEAKER_01: Now it's cobot, meaning that collaboration of the robot and humans, and they need to work shoulder shoulder type of thing.
• 47:51
  SPEAKER_01: So you need to have a lot of awareness what's going on around you.
• 47:54
  SPEAKER_01: You need to have a lot of sensing modalities, you need to have a lot of edge processing because sometimes robot needs to react to something uh very quickly.
• 48:03
  SPEAKER_01: You cannot send it to cloud and wait for cloud to to kind of analyze it and send you uh back to you what you need.
• 48:09
  SPEAKER_01: So edge AI becomes important, sensing becomes important, motor drive and all kinds of capabilities to make the robot more intelligent becomes uh very critical.
• 48:20
  SPEAKER_01: And now, I mean, again, at some point robots started showing up in assembly lines.
• 48:26
  SPEAKER_01: Okay, close to human, but still kind of repetitive work.
• 48:29
  SPEAKER_01: Yeah, but now we are bringing them to the point that they uh have some autonomies that, like, you know, they can make a decision and move around.
• 48:37
  SPEAKER_01: So that pushes all of these pieces that I mentioned to the next level of technical challenges.
• 48:43
  SPEAKER_01: That, you know, now your sensor needs to be much more reliable and much more uh aware of the surrounding.
• 48:51
  SPEAKER_01: Your motor drive needs to be a lot more efficient because these things are running off of a battery.
• 48:55
  SPEAKER_01: They are mobile, they are not plugged in, unlike these big robots and stationary robots.
• 49:00
  SPEAKER_01: Uh, they need to have the local inference and processing.
• 49:04
  SPEAKER_01: So we are working on all of these and making sure that you know they fit to the kind of processing and power budget and cost budget that uh a drone or a robot or a kind of uh autonomous system requires.
• 49:20
  SPEAKER_01: So it's very interesting how these vectors intersect in different markets in a different way and pushes the technology to the limit.
• 49:28
  SPEAKER_00: Well, it's been absolutely fascinating, Dr.
• 49:30
  SPEAKER_00: Baha'i.
• 49:31
  SPEAKER_00: Thank you so much.
• 49:32
  SPEAKER_01: Thank you.
• 49:33
  SPEAKER_01: And good talking to you.
• 49:35
  SPEAKER_00: Thank you so much to Ahmad Baha'i and Texas Instruments.
• 49:38
  SPEAKER_00: Thanks to you, as always, for staying until the bitter end of the program.
• 49:42
  SPEAKER_00: Please like and subscribe on the platform of your choosing.
• 49:47
  SPEAKER_00: And don't forget to check out the newsletter and more over at automated.fm.
• 49:51
  SPEAKER_00: That's all we got for you this week.
• 49:54
  SPEAKER_00: We'll see you next week for another episode of Automated.