Exclusive / Applied Intuition wants to turn robotics into child’s play

Applied Intuition’s Sunnyvale headquarters are unlike most Silicon Valley startups. Cars, trucks, and big rigs sit in various states of disassembly as engineers test and validate software designed to turn otherwise ordinary machines into rolling robots.

In one section of an enormous garage, a pickup truck is being programmed to enter the battlefield, without a driver, and launch anti-drone munitions. Around the corner, a self-driving truck is a guinea pig for autonomous driving software destined for Japanese roads.

In the offices above the garage, employees work in slippers at spotless desks — a mandate inspired by Japanese culture that Younis picked up while working for Bosch in Japan.

Younis, who cut his teeth at General Motors in Detroit before beginning his Silicon Valley career, saw an opportunity in helping traditional automakers catch up to the tech industry on software. Applied Intuition, for instance, is building the operating systems that run everything from infotainment systems to automatic doors.

The company has since diversified into other industries like construction, agriculture and defense. The automotive industry, he says, accounts for roughly a third of the company’s revenue today.

Along the way, Applied Intuition has collected a decade’s worth of data and expertise on imbuing most things that move with “physical intelligence.”

Peter Ludwig, Applied Intuition’s co-founder and chief technology officer, talks about entirely new categories. “Imagine if you could buy a robot to maintain your garden. There’s going to be a lot of people right now that just have unmaintained gardens that are going to buy that thing because they want a nice garden,” he said.

There is also a missing-link effect in robotics: Every gap that gets automated creates demand to automate the ones around it. In a biotech lab, robotic pipetting machines can already run experiments, but humans still have to set them up and ferry samples from station to station. A robot that fills that gap turns it into a facility that runs around the clock.

Younis sees a repeat of the smartphone era, when years of infrastructure-building suddenly gave way to an explosion of applications: The universe of physical AI applications, he added, “really will make the App Store look small.”

One example is the construction industry. In the US especially, construction costs have skyrocketed and the labor force has dwindled, slowing the development of infrastructure and new housing. “We would build a much bigger solar farm, we would build many more windmills, effectively, if we had the labor to do it,” Ludwig said.

Dana looks less like the industrial software it replaces and more like a chatbot. In a demo, an Applied Intuition product manager typed casual requests into a chat window: “Show me a segment set.” The system pulled up camera footage from the company’s data-collection cars in Stuttgart, Tokyo and San Francisco, with labeled data that companies can use to train or finetune their own models.

“There are parts of our platform that you almost need a PhD just to understand,” Ludwig said. Dana puts an easy-to-understand AI agent on top of all of it.

The difference shows up in tasks like validating automatic emergency braking. Automakers used to assign systems engineers to read the entire New Car Assessment Program handbook, translate dozens of requirements into test scenarios by hand, and run them. Now Dana runs the simulations and returns a graded report on how the braking system performs, along with the specific situations where the car fails to stop in time and suggestions on what to fix.

Asked for a snowy day, Dana’s world models regenerate an entire driving scene in snow, rain or sunshine, across every camera on the car. Simulated traffic is populated by AI drivers trained through reinforcement learning to be deliberately difficult, cutting off the test vehicle to probe its weaknesses.

Applied Intuition has been using Dana internally for over four months and with select customers like Isuzu and heavy-equipment manufacturer Komatsu.

Marc Andreessen (who happens to one of Applied Intuition’s major investors) once wrote that software was eating the world. He was right, but software could only reprogram so much. When we have software that can see, understand and manipulate its surroundings reliably, the physical world gets eaten, too.

We’re obviously not there yet, and it may require robots with hands as dextrous as the human versions to fully complete the transition. But I don’t think people have fully digested how profound this change will be.

These advancements are often looked at through the lens of labor replacement, but that is extremely narrow. It will be more like 3D printing reality itself.