Tesla’s former head of artificial intelligence, Andrej Karpathy, who worked on the automaker’s self-driving effort until 2022, warns against believing that self-driving is solved, and fully autonomous vehicles are happening soon.
Karpathy is a very respected leader in the field of artificial intelligence.
In 2017, Musk poached him from OpenAI and he quickly became the head of Tesla’s AI effort, including leading neural nets for Autopilot and Full Self-Driving.
He left Tesla in 2022 and return briefly to OpenAI in 2023 before starting his own in AI education company, Eureka Labs.
The Slovak-Canadian computer scientist is widely regarded as one of the top computer vision experts and he pioneered Tesla’s vision-only approach to self-driving.
Karpathy gave a talk at Y Combinator’s AI Startup School event this week and made some interesting comments about self-driving.
He recounted when a friend working at then Google self-driving company, now Waymo, gave him a ride in a self-driving car in 2013:
We got into this car and we went for an about 30-minute drive around Palo Alto, highways, streets and so on, and that drive was perfect. Zero intervention. And this was 2013. It was about 12 years ago.It kind of struck me because at the time when I had this perfect drive, this perfect demo, I thought “well, self-driving is imminent because this just work. This is incredible.” But here we are, 12 years later, and we are still working on autonomy. We are still working on driving (AI) agents. Even now, we haven’t actually solved the problem.
12 years later, Waymo currently operates over 1,000 vehicles in California, Arizona, and Texas where it completes hundreds of thousands of autonomous rides with paying customers every week, but Karpathy explains that this doesn’t mean autonomy is solved.
He continues:
You may sees Waymos going around and they look driverless, but there’s still a lot of teleoperation and lot of humans in the loop in this driving.
Waymo has confirmed that it uses some teleopeartion, but it’s not clear to what level. It’s clear that it at least communicates commands to the vehicles remotely when they get stuck.
Kaparthy adds:
We still haven’t declared success, but I think it’s definitely going to succeed at this point, but it just took a long time.
The engineer added that “software is tricky” and that he believes that “AI agents”, which is a term often use to describe AIs that can perform tasks for humans, like driving a vehicle, are going to take time. He believes this is not the year of AI agents, but the decade of AI agents.
Here’s the full presentation:
Electrek’s Take
While Kaparthy didn’t name Tesla, the timing of his comments as Tesla is launching its “Robotaxi” service this weekend is interesting.
It certainly contracdits what his former boss, Elon Musk, is saying: that self-driving is solved.
He highlights the fact that humans are still “in the loop” in Waymo’s vehicles, but we recently learn that this is even more true with Tesla’s Robotaxi launch, which involved not only teleoperation like Waymo, but there’s also a Tesla employee in the front passenger seat ready to press a kill switch.
As we have often highlighted in recent weeks, Tesla’s Robotaxi launch is simply a game of optics for Tesla to be able to claim a win in self-driving after years of broken promises and missed deadlines just as Waymo is rapidly expanding its own self-driving services.
I think Kaparthy, who led Tesla’s computer vision effort behind self-driving, knows that has yet to solve the problem and will require human supervision for a while longer.
Based on the best data available, Tesla currently achieves a few hundred miles between critical disengagement with FSD and it needs to get into tends of thousands of miles to achieve a true level 4 autonomous systems.
We are still a few years away from that at best.
FTC: We use income earning auto affiliate links. More.