Pickle Robot Company

Trust, collaboration, and scalability are shaping the future of automation. In the latest episode of Code the Future, Ariana Eisenstein, Founder & CTO of Pickle Robot Company, joins Matt Chandler of QNX, a division of BlackBerry, to discuss the realities of bringing robotics into the workplace, building trust in autonomous systems, and creating meaningful human-robot partnerships. 🎧 Don't miss this insightful conversation: https://lnkd.in/eqarDib7 

There are only 3 ways to get intelligence into a robot. 1️⃣ The first is advanced modeling. You hire PhDs to write down the equations and solve the physics analytically. That's optimal control. 2️⃣ The second is simulation. Millions of hours of it, running reinforcement learning. The problem is deformable objects. Cardboard and bags look simple. They're really complicated, and our simulators don't capture that. 3️⃣ The third is demonstration. A human teleoperates the robot and shows it what to do. On their own, none of the 3 adds up to a solution for the hard problems our customers have, like loading and unloading trucks. So the Dill Autonomy Engine uses all 3 at once. At the center is the model control sandwich: guidance and skill agents on the outside, a diffusion-policy generative AI on the inside. Around it, sensors, observer and supervisor agents, and a teleoperation interface route on-policy data back into training. The robot learns the basics from modeling and simulation. It masters the job by doing the job. It takes very little real-world experience to get there, and that's why I think Pickle is right at the frontier of getting robots to use their hands productively. Part 3 of my series on how we built DAE. Full video below.

The Dill Autonomy Engine is the brain behind our robots. I built Pickle Robot Company to solve one job, truck unloading. DAE is turning out to be the answer to a lot more than that. More on the hybrid AI architecture behind every Pickle Robot: https://lnkd.in/evixFx7W #PhysicalAI #Robotics

We started Pickle Robot Company on a bet: automation was too expensive, so robotics wasn't touching the jobs that actually needed it. Early on, we walked a parcel sorting plant at 2am. Fifty people loading and unloading trucks by hand. Squishy boxes, wet cardboard, freight that falls apart the second you touch it. Physically one of the hardest jobs I've ever seen. I told my co-founders: if you can get a robot to do that, you can get a robot to do almost anything. It's also a real business from day one. Load and unload trucks well, and you generate revenue immediately. The better the robot works, the more value it creates for customers. The more you sell, the more you reinvest in R&D, and the better the robot gets. That flywheel is a far better way to build technology than writing grant applications. That's the bet we made. It's still the one we're making. Proud to announce the Dill Autonomy Engine to the world. I’ll be posting several videos in the coming weeks to expand on this - stay tuned! More on why we built the Dill Autonomy Engine: https://lnkd.in/evixFx7W #PhysicalAI #Robotics #PickleRobot