Executive Officer, Tesla, Inc.
AI World in 2035: Tesla VP Ashok Elluswamy’s Bold Prediction | Gobinath
Enquiries:
[email protected]
Stay ahead with direct updates from Gobinath — interviews, insights, and powerful voices.
Join the official WhatsApp Channel today:
https://whatsapp.com/channel/0029Vb5g...
#Gobinath #ashokelluswamy #tesla #elonmusk #ai #artificialintelligence #ev #evcars #roboticsengineering #podcast #agi #artificialgeneralintelligence #humanoidrobot #humanintelligence #adas #quantumcomputer #software #hardware
In this groundbreaking episode, Gobinath sits down with Ashok Elluswamy, Vice President of AI Software at Tesla, for his first-ever in-depth interview. A proud Ta...
About Ashok Elluswamy
Ashok Elluswamy, Vice President of AI at Tesla, has described the company’s mission as “producing amazing abundance for the entire world” through autonomous vehicles and humanoid robots. In early 2026, he stated that Tesla’s robotaxi service is operating in Austin, Texas, with no safety driver, and that the company aims to scale the service to the entire fleet. Elluswamy said that Tesla’s Full Self-Driving (FSD) system is an end-to-end neural network that uses only cameras and no radar, lidar, or HD maps, and that the same foundational model is being adapted for the Optimus humanoid robot. He predicted that by 2035 all cars manufactured will be self-driving and that humanoid robots will be able to interact in natural language. Elluswamy has emphasized that Tesla’s approach relies on large-scale data and compute, including the custom Dojo training hardware, to achieve safety and smooth driving. He stated that the company’s end-to-end system can be probed for reasoning traces and 3D understanding, and that it has enabled features such as emergency braking for crossing vehicles. Elluswamy described Tesla as “all in on robotics” and said the company is focused on producing intelligent robots for physical work, with the goal of creating a car that “doesn’t need to crash ever.”
Transcript (40 segments)
N
Narrator0:04
Elon Musk expressed his appreciation for Indian-origin robotics engineer Ashok Elluswamy. Ashok is actually the head of autopilot engineering.
A
Ashok Elluswamy2:33
Hi everyone, my name is Ashok. I lead the planning and controls, autopilot, and simulation team.
I
Interviewer3:07
I'm really excited to do this interview.
A
Ashok Elluswamy3:36
Tesla Vice President. Hello. Nice to meet you. Nice meeting you. Thanks a lot.
Yeah, of course. No. Um, for the most part, four years. Then I switched to an automotive company. But I was always interested in robotics. So I was doing a lot of projects on robotics and I wanted to study more in robotics.
Computer. It was giving so much pleasure and like so fun. So then it became so interesting to me. Particular, I was like thinking about it all the time, but everyone thought like robotics was not ready as a field back then and don't study robotics. I didn't care about it actually because it was like so fun. Like 15 years ago, I was reading what are the future of robotics and so on, and I realized it was robots. I was pursuing robotics. I wanted to work on something closer to software that moves hardware.
Software, I think it's quite hard technically. It is quite challenging. Hardware, you have to physically build things which requires like good expertise in manufacturing, like motor design, mechanical design. It's not as scalable in some sense. You can infinitely clone it and then distribute it. So there is inherent scalability in software. So it's very attractive. So you can get a lot of reward for it. People are going behind it. But applications.
So probably like you know, 50 years, 200 years, there was an industrial revolution where like machines started doing it and then there was a tremendous boom. So people went all in on mechanical, like you know, clothing. We went behind machines, made machines do the hard work. We are not doing the mechanical work anymore. Machines were just the muscle and the software was just the brain. Now we are putting both of them together into single. Okay. So muscle and brain when you put together, next wave of physical progression, it is coming through robotics. It can do more tasks on its own. It is trapped inside a computer. Muscle and brain.
In the future, next 10 to 20 years, there's going to be a tremendous wave in robotics. 2012 landmark paper, graphics processing units. More, it was able to beat the previous performance by 25 percentage points. Like it was a big improvement, nearly 1% improvement, they improved by a very large margin. Okay. Wow. They started changing. So 2012, continuous compounding, every year there were breakthroughs. So perfection, perfection, and new architectures. 2017 language. 2017, it became so good at that task. Leading models. It is based on very similar architectures of combination of convolutions, transformers, deep and 2012 exponential grow. And next 10 years, we will have artificial general intelligence. We'll definitely have that in the sense that humans and AI-based software will also be able to do the same thing. Probably within 15 years, even athletic professions, like robots will be able to play soccer, cricket too.
Importance. What I think it is progress. It's not likely. So it is strictly getting better. Strictly getting better.
Yeah. I since I work inside it and many people who work inside, this algorithm is not going to stop. You can make it bigger, add more data. You can continue to scale it up. No one is seeing why it shouldn't work even better. Okay.
Since you are there, but since you are there, EV, okay, not just AI. So the dream world, 2035, all cars manufactured will be self-driving. If you want like specialized hobby, you can get older cars.
We are expecting to launch driverless in some cities in the US. Okay. Actually. Yeah. So, which is humanoid robots. It will be able to do industrial tasks, like also inside the house. Domestic work. And you can interact with it in natural language. Yeah.
Okay. Technically possible. Yeah. So we have to make decisions around like robot, it's a moral question also. Okay.
Big. All research has shown that person. If you give dedicated attention instead of a classroom setting, okay, they will grow even better, like way better, like two to three times learning than classroom setting. Of time, they're not able to follow. So it's not affordable, right? You can actually individualize education. Every person as a personalized teacher who understands the child and delivers education for them.
Movie direction and traditional graphics animation. You can just describe, you can point and interactive design. As a single person, you can create maybe an entire movie on your own without any production crew, actors, no one required. As a single person, mind, if you have a story that you want to communicate and visually, if you have clear idea, you can direct an entire movie by a single person with the help of AI generation. New jobs will be created. And overall, we are trending towards abundance.
Abundance. Poverty, no. People are still having like a lot of people work for passion, but a lot of other people also just work for literally just money, just earning. And they are doing this work just to get that means. But why should they even do that? It's a waste of the creative potential. Why are they going to work? It will be a future where there's abundance. Maybe not in 10 years, but 50 years. I think definitely fundamental. Okay.
So, the automatic muscle and the brain together of machines. It's my future. Obviously my version of the future, but your version of future, putting together your knowledge, your assumptions, your analysis of future. Yeah, future, obviously I believe this is going to be, this is going to happen. Okay.
Using all the opportunities. I think it will make the tools available for everyone. Powerful tools. They have tons more resources. They able to read everything already, a lot more resources. They can have like very high success or just live normally, long capability. But as agency, agency, billion, thousand people, 100 people, 10 people, let's say one billion, even one person, they could do it. Then you have the resources.
Yeah. Okay. Advanced driver assistance system. You need way more reliability to remove the driver. 1, 2, 3, 4, 5. Base driver has to pay attention. Okay. System. Full safe operation, very undefined area restrictions five. So four and five almost same, same gray area, it's not very well defined.
Artificial based city road. So intersection traffic, traffic lights, stop first intersection. 360 degrees, very powerful computer networks running on board in the car. Just like how a human would perceive steering and acceleration values.
It's not just programmed. It's not programmed. Okay. It is learning, machine learning. Okay. Completely learning process. Learning process.
Rest top humans, race car drivers, 130, 150 delay. Okay. Okay. Doing magic. Again. 2035. 20, for example, continuous progression of technology. Early driver, it's a combination of the AI software and the driver together system. There's a lot of driver monitoring to make. We have a lot more driver monitoring technology.
It will make sure the driver is engaged in the task. Technology and future technology. We don't want any confusion. So the driver, they have to be there. And future driver, our safety will be high enough. We don't need the supervision of the driver. We will handle everything safely. And if anything happens, the car is responsible. We are going towards a world like that. And there's already like some driverless vehicles in the US, like there's already a service called Waymo which is operating driverlessly, which is already safer than human driving. Yeah, it is. It is operating artificial intelligence and driving. It's impossible.
Network. It's basically a digital living being, human being. Mhm. It's not like which is more, which is obviously going to be more intelligent than any other animal. Obviously, we can communicate with that. Okay. And even though it is AI, it's the same AI in every car. It is not like every car has its own intelligence.
Human artificial middle artificial intelligence. Normal situation. It will make progress. We don't want angry software. Like, yeah. Trending towards, it will build towards yielding and safe driving behaviors. Not with sufficient level of, without sacrificing safety. Clear balance between, always choose towards somewhat conservative driving behavior.
Artificial buy any car, software, you can go in there and then address, you can just press. But you don't have to drive. It should respond automatically. Comfortable. I'm not sure it's responding correctly. Sometimes the people take over and manual driver assist. Yeah. Driver supervision completely artificial intelligence technology. Of course.
But why I think progress, deep learning, massive, last it has been even more insane. So it comes down to computer sort of speed. So you can't make the algorithms larger and like make the model bigger. Back there have become fast enough that you can afford to do the same thing. The core technology was invented even earlier. They were able to parallel compute, but they were able to creatively use it for this one and found that okay, it's working much better. So then once they figure it out, investing heavily in both the software but also the hardware. Hardware is also critical part of it. Okay. Heavily investing, a lot heavily investing on a technology always kill one technology.
Artificial intermediate arrangement. I think quantum compute, it's only, it's challenging the hardware that's running AI as opposed to the software itself. It's how the brain works. We can go to different computer. It's a meat computer. It's based on chemicals. Digital computer is already not like that. It's already different computer. But the algorithm is pretty similar. Not exactly the same. We don't fully understand how the human brain works. But it is very similar to how human brain works. And they are flapping their wings and flying. It is the fundamental physics principles are the same. The mechanical actuation is different. The fundamental principles are the same. Implementation, it's different. It could be, it could look, yeah, but the fundamental concepts of what drives flight, it is similar. Okay. Force, fundamental physics concept. Implementation can be different forms.
But yeah, the same software. The same software. For example, 10 GB, 100 GB. Google. 2012 robotics. It was very hard to make progress. It's blind. They had to make some progress. Okay. They were then tending towards those technologies where depth director. So then you can make some progress on robotics and they were able to make progress. 2009 technology competition camera mostly and it started out from the same technology and they are continuing down that path. The main disadvantage of LiDAR, quite expensive and it is an active power emitting device as opposed to camera, passive device. It has to emit the light. Camera is illumination from outside. So it's very power efficient camera.
When it started out, it was like $75,000, $50,000. Few, so cheap, mobile phone, you can get cameras for $10, $5. In the end, we want to produce scalable autonomy that's affordable by everyone. You have to bring down the cost. For example, 100, same comfort, same safety. Obviously, you need a lot of it is technically challenging to produce that configuration. It is more valuable for the world to have a very low cost solution to this problem. So, so I mean, equal quality, equal quality. Technically, it is already performing. We are like maybe like, you know, like maybe lagging by couple years. We would have been forced to use maybe some other technology, but now technology, it is possible and we decided to go with that strategy. Okay.
But still, it was clear that the ceiling of what I can achieve was very limited and traditional, like jump through the hoops. It was very depressing to me. My view and view, people's view. It should be based on talent, not based on age of the person or like, second, it was not based on some scalable software solution or something like that. It was based on, not like, it was not like super innovative area. It was a terrible decision. He was working on like some plant. Oh, okay. Housewife, home assistant. Yeah. Yeah. Okay. Still. Oh, okay. Idea.
Last venture. Not like public school. So I was focusing on that kind of thing and I was able to get very good score in entrance exam. I was able to join College of Engineering. College. I took ECE and I was, computer gone to software and throughout my bachelors also I was just secretly like doing software projects but for robotics. So the combination of electronics and software, college basics, but I went to like, I or I go there and participate and so on.
Technology. Wife, she changed recently but she used to work in Amazon as a software person for AWS cloud service trading app is like so valuable for everybody. Okay. Yeah. Yeah. And we got married and we have been together for like 11 years now. So cool. Nice. Of course.
Is the brain very calculative or a fantasy? Brings that. Yeah. He's really smart in the sense that way earlier than other people, which is a very big gift for anyone to have. And he works really hard, like easily 80, 90 hours per week for like multiple, like many, many years through his entire career. So he's really hardworking and talented. So I feel fortunate to work for him, work with him, and so on. Like he's not afraid of taking risk. Ah, again, I think I naturally was able to like survive and flourish that probably also because of some similar character traits where like he's very funny, mindset, solution, and very straightforward. Great person to work with day-to-day. But fortunate to be in this position, like ended up here. So I'm trying to do the best possible from this position.
I think it's a bit of both. They're not sharing all the secrets. But they are sharing some information and it's like half collaboration, half competitive. Okay. Overall philosophy, philosophy, every American should have a car. Philosophy. But overall, what's the philosophy of Tesla? Philosophy is sustainable abundance. Sustainable abundance. Yeah. Abundance. Fossil fuels. It might not expire in 10 years but you know, 100 years it might expire. So it is not sustainable by the definition of it. Sustainable in the sense that it is not a sacrifice for something else. Okay. It is, it's a win situation for the world and for the person who consumes it. It is a saving mindset, but you can produce more. Okay. It is the mission itself. Yeah, of course. Sustainability, abundance, but still, yeah, it's possible to capture energy for basically, energy is basically free. We can create abundance by combining machine with smartness.
Okay. 10 to 12 hours to. Okay. Yeah. You only have two priorities. Work and family. Work and family. Sometime obviously like meet friends and external family and so on, but mostly work and kids and simple. Yeah.
Okay. Okay. And again back to this thing. For they will be going in cars. It's not worth the trouble basically. It is premium experience. So why would you go? It's like strictly painful to own something in the process. Yeah. So, we have to do infrastructure both for like the car and the roadways and the other things. Welcome to New York. Cost of transport. So more people will be able to use cars. Yeah. Traffic will anything increase but we have to improve the infrastructure. Even here, metro tunnels, they're able to like build additional roadway capacity effectively by drilling holes under the ground which is quite safe. So tunnels is a good option and the traffic, there will be more traffic. We have to like come up with innovative solution like tunnels solution and caring more, more space. So parking lots are not necessary so you can use that space for parking lot. Yes.
How Indian car industry will grow next? Of course, especially artificial intelligence plus EV companies. I'm not the expert in Indian market to be honest. US cities, they like quite far away and you need lot of range to cover those cities and so on. India, the distance between cities is much lower. It's similar to European cities and so on. India, electric cars are more amenable to the travel patterns that typically there in India. You can still go on longer trips but their range that required in US is much longer. So the cars are more expensive there. You can get cheaper electric cars because the range, battery is one of the major components of cost. So, second, I think driving, the companies are working on it. Yeah. And it will reduce the fatigue of driving. Like it's not going to drain you. It'll be so tiring just from the driving and it's just like the same menial driving. It'll take the chore out of driving. You can just supervise and have a relaxed driving experience.
Engineers in India, core concepts, there's a lot of resources online to read and study and like experiment models. Next, not, they shouldn't target four years. What is required for that, they have to plan for, we have to be prepared for that. So, happy is all mine. I'm so happy. Technical conversation opportunity. Thanks again. Best wishes. Will come to factory times. We will catch up there. I wanted to come and visit how this thing, technology works. Definitely. Yeah. Okay. Thank you. Thank you. Thanks a lot. Thank you. Thank you.