What I Learned Building AI That Competes with TeslaㅣHelm.ai, Vlad Voroninski

00:00The end goal is achieving fully autonomous driving. We allow automakers to compete with Tesla by bringing to market cutting edge autonomous driving systems. We've raised over $und00 million. We'll see very very scalable commercialization happen.

00:13But I would say that starting a company felt like drinking from a fire hose. When we were pitching kind of like seemed like a pipe dream, pure R&D for 2 years. There was zero product development during that time. Everyone thought it was crazy.

00:25But we committed to that vision. We actually carried out the research and we were able to make that work very much all or nothing and it might take a long time for the world to adjust. My name is Li Berninski. I'm the CEO of Helm AI. Helm is an AI software company focusing on a unified approach to autonomous driving that goes all the way from L2 plus through fully autonomous driving L4.

00:48We have partnerships with companies like Honda and we allow automakers to compete with Tesla by bringing to market cutting edge autonomous driving systems. The end goal is achieving fully autonomous driving and there are a number of technological and commercialization challenges along the way.

01:05These days, one of the key areas that we're focusing on is AI based simulation, leveraging generative AI as well as our unsupervised learning IP in order to create a unified approach to solving autonomous driving that essentially unifies both partial automation and full automation.

01:26First got interested in self-driving cars and computer vision during undergrad. So, I was part of the UCLA computer vision lab while they were competing in our grand challenges. We're in the offhighway vehicle recreation area. So we just did our first autonomous path tracking test and I just thought that was a very exciting area and was focusing on computer vision at the time but decided to pursue mathematics in academia for about 10 years with the intent to come back to the space when it was more mature because I saw that as the key bottleneck to AI in the sense

02:06that the biggest challenge in in reading research papers in AI or computer vision was ultimately a mathematical you know just a question of do you understand sort of the equations right so yeah in some sense I looked at math as a tool to be able to solve AI down the road the goal was always to come back to the autonomous driving space in the war between humans and artificial intelligence this is for 33-year-old professional go gamer Lisa D Elon Musk he plans to equip all new Tesla cars with the hardware needed for full self-driving capacity around I would say

02:392016 it became clear that the technology was really taking off in terms of deep learning. There was an inflection point in where AI technology was going with uh deep learning at the time and simultaneously there was a very clear opportunity that it was the right time to jump into that space because what I witnessed was a lot of companies making certain decisions that I actually didn't agree with.

03:03Right? It was such an inefficient space at the time that it was very clear to me that with the right approach you can add a lot of value. Right? because uh it's a strategy that's kind of stood the test of time as opposed to many companies that peaked early and then died off or just kind of ran out of money and what that meant was that there was an opportunity.

03:22So after my academic career moved back to California to start Helm Co was a very kind of a tricky time for everyone obviously but also in in the automotive market in particular because it basically caused a halt in production. The Corona virus is idling one auto plant after another.

03:43All the different auto factories and all the automakers had to immediately start dealing with that issue versus everything else. And I think it caused a bit of a delay in the deployment of autonomous driving technology. But beyond that, I mean, I would say quite exciting.

04:00I mean, ultimately, I I I don't know if it was like I think the challenges were there, but they like were outweighed by how exciting it was to like start a company, you know, make a truly uh deep tech bet in the space. Our first 10 hires were basically all just very, very strong researchers.

04:15Any one of those people could have easily walked away and done something else. Some of those people even made certain sacrifices of their academic career to come work at Helm because they were very excited about the vision. And I think that helped us really mold the engineering culture.

04:32When we were pitching Helm 2016, right, unsupervised learning was kind of like seemed like a pipe dream almost, right? Um, but we committed to that vision. We actually carried out the research and we were able to make that work. So that was uh, you know, quite exciting.

04:48So I mean I guess like there was the fact that you know for 2 years we were essentially developing that technology and there was zero product development during that time. So it's pure R&D for 2 years very much all or nothing. Um so you know uh obviously there's risk involved in that but it was a very creative time so I I mostly just appreciate it.

05:14an Uber self-driving vehicle that flipped. Uber is now banned from testing its self-driving cars in Arizona. Tonight, Tesla confirming this car was in autopilot mode when it crashed in Northern California back in 2018. Kind of foray into foundation models, even before that term was coined.

05:35What we used that foundation model for was to essentially build an autonomous driving system that we were able to show actually outperforms the systems you were able to buy in the market by pretty wide margin. So we essentially conducted a series of tests where we put our autonomous vehicle on uh very steep and curvy mountain road scenarios where essentially you have to make very rapid driving decisions as far as you know taking the various turns in a challenging uh landscape.

06:01And we were able to achieve much better disengagement rates uh up to a factor of 200 better than what was out there on the market. And that is how we got the attention of some of the brand name automakers in the world in the early days. In the last couple of years, we've been doing a lot of innovation in generative AI and combining that with our in-house technology which is called deep teaching in order to close the gap between AI based simulation and reality.

06:31So essentially that means how do you simulate driving data or driving footage sensor data from driving without actually having to get into a car. And there are many advantages to doing that. For example, very large fleets, right? They can be useful for collecting data in order to address difficult corner cases for autonomous driving, but the rate of occurrence of those corner cases basically goes down exponentially as your system improves.

06:54And so end up actually paying exponentially more to gather interesting data as you get further into the development process. So it's really not a good property. And what simulation allows you to do is generate all the interesting data without actually having to deploy a fleet.

07:10For example, Tesla that has a very large fleet. Other automakers don't have access necessarily to internal fleets that are that large. So even if they wanted to take the same approach as Tesla, they would not be positioned to do so. The only alternative to doing that is essentially AI based simulation.

07:27And until very recently, it wasn't possible to generate highly realistic simulation data. But that's very much changing these days due to the advent of generative AI and combining generative AI with technologies like deep teaching provides a highly scalable uh simulation platform that allows you to essentially deploy a virtual fleet so to say instead of a real world fleet you're just uh learning from existing data.

07:50So that's a recent inflection point that we're definitely you know proud to be part of and contributing to. Vid Gen 1 and World Gen 1 are foundation models for generative AI simulation. Vid Gen is a foundation model that creates highly realistic video data from a multitude of different cameras, essentially arbitrary cameras, arbitrary locations.

08:11Worldgen is a foundation model that takes a further step in that it actually simulates the entire autonomous driving stack. You can actually use Worldgen to technically to drive a car because it does make predictions about what's going to happen next.

08:26So if you input data from your autonomous driving stack, it'll tell you what's going to happen in the next several seconds and that includes the path that the vehicle should take to perform certain actions. You can actually use it to drive.

08:38So it's technically uh you know a self-driving system that not only functions as a simulator or in a simulator environment, it also can function in the real world. Yeah, basically how to actually convince a customer, how to develop a relationship with a customer in autonomous driving space.

08:58Um, I think there's kind of two important aspects at least. So, one I think is, you know, seeing is believing. Not only having kind of marketing materials or video demos, right? actually being able to put somebody in a car and have that autonomous car navigate unforeseeable situations um I think is a very powerful sends a very powerful message about the robustness of the technology and the product and secondly I think that working on a production contract there are going to be kind of other contracts along the way right I think it's not really possible for a major car company

09:31to give a production contract to a supplier as the first contract so there's going to be some sequence of contracts along the way and so I think that being able to execute on those contracts and deliver exactly what you signed up to deliver is critical because ultimately what you're entrusted with as a supplier is providing not only safety critical technology but technology that is absolutely necessary to have in a certain timeline because you're talking about a production program that has to launch in a particular year where you know a lot of money is a lot more money

10:03is invested into that than just the money being paid to one supplier. So, it's incredibly important to be able to meet those deadlines. Maybe a third thing I would add is just not only demonstrating your current state of where your technology is, but demonstrating the the difference from one time to another, right?

10:20So, being able to show, okay, here's where we are at this point in time. And then in a month, we expect to be here, right? and actually showing, you know, showing that differential and allowing them to measure not only kind of your position, but also your velocity, so to say.

10:34When it comes to technology development, especially when you're going after such an ambitious play like autonomous driving, kind of requires quite a lot of conviction. The hardest part about it was just if you're a researcher, I think that your key job is really okay to to perform the research and then you put it out there.

10:54And sure, there is some marketing aspect to that, but it's not nearly as significant, I think, as what you have to do for a company. And generally, as a startup founder, you have to wear so many different hats. People say when they go to MIT, for example, that it's kind of like drinking from a fire hose.

11:10I never had that experience, but I would say that starting a company felt like drinking from a fire hose. You know, in our particular case, I think we were in some sense working against the grain in that the vast majority of the funding went toward companies taking a totally different approach in that they were pursuing pure play 4.

11:28And you know, when I first started to engage people about the fact that we're going to really focus on partial automation as the key market, you know, everyone thought it was crazy. But I I think that it basically just emphasizes more kind of this this notion of the importance of like having grit or something because you can't expect what you think to be the predominant world view and it might take a long time for the the world to adjust right like I think we only started seeing signs of the world adjusting to our point of view some number of years into the company right

12:03maybe four years in 5 years in right but then every year the our position has improved prod not only as a function of the technology and the product but also because our strategy was adapted to a certain worldview that we believed would essentially would eventually materialize and that is now happening.

12:24[Music]