While many people focus on the negative connotations of robots becoming self-aware and self-replicating, the field is still very interesting for those exploring the world of creativity, artistry, and imagination for robotics and artificial intelligence.
One of the leaders in this space is Dr. Hod Lipson, the James and Sally Scapa Professor of Innovation of Mechanical Engineering and Data Science at Columbia University. Lipson directs the Creative Machines Lab, which pioneers new ways for novel autonomous systems to design and make other machines, based on biological concepts.
As part of his journey in robotics, Lipson has co-founded four companies, co-authored more than 300 publications, and co-authored books on 3D printing and driverless cars. His TED talk on self-aware machines is one of the most viewed presentations about AI and robotics.
Joanne Pransky, associate editor for Industrial Robot Journal, recently spoke with Lipson about his career, his successes and failures, and the surprises he discovers from self-replicating robots and creative machines.
The full interview is available free to Robotics Business Review readers until December 25. Here is an excerpt:
The early days
Q: What led you to co-found your first company, Tri-logical Technologies?
Lipson: I founded Tri-logical in 1994 before I started my Ph.D. I was living back in Israel, where forming a startup was a rite of passage. The question really was, “What was your startup going to be about?”
Even back then, Global Positioning System (GPS) sensing was not very accurate, but it was becoming cheap and portable and there were many things you could do with it. We had a lot of fun trying to imagine a future where you could know where you are. The big question was what could we do with GPS that was only accurate to about 100 meters? We figured out that the one really useful thing would be to track shipments, i.e., to know when something you’re shipping would arrive from across the country, if it’s on track, if it’s deviating from its route, etc.; all these things were very well suitable for 100-meter resolution. That’s what we developed and sold it to insurance companies. It ended up being very successful. It also highlighted for me my interest in academia, and in moving away from product development towards more basic research. So I left that company and started to do something else to facilitate other research efforts.
Q: Can you tell us about the successful applications for the VERSABALL gripper, the flagship product of Empire Robotics, one of your lab spinouts from Cornell University?
Lipson: The Jamming Gripper was an ingenious invention, but it didn’t sell enough to survive. And that was another lesson – that it’s not just the technology, but it’s also the business model around it. You can have the greatest technology but you need to create a market and sustainable business plan around it.
Q: What’s your favorite creative machine?
Lipson: One of my personal favorites is my brainchild, robotic painter Pix18. It creates original, generative art on canvas (see photo, above). It lives in our living room. One of the newest things that it’s creating are portraits of people that don’t exist; people that it makes up.
Pix18 is walking around the world right now in Google Street View and it will pick something out of this experience and paint something that it has seen and liked, in the form of neural activity. The robot has experiences that I don’t have. It’s an example of how AI in the past five or six years has moved forward by leaps and bounds, doing many things that were impossible just a few years ago, surprising even people in the AI business.
Q: You build robots, as you’ve said, ‘that do what you’d least expect robots to do.’ What robot behavior gave you the least expected response?
Lipson: Making machines self-aware is my career Holy Grail. I would say that the robot that was most surprising to me is the self-replicating robot. We had one project where we created a universe of robot pieces in simulation, with no explicit goal or motivation. We sat back and watched. What happened in that simulation is that when we left it running long enough, a self-replicating robot emerged – spontaneously. Self-replication is the ultimate goal.
That was a huge surprise to me, one that you’d least expect. The robot did not want to self-replicate, but just statistically speaking, replicating robots dominated very quickly over those that did not self-replicate. We just let it find its own mutation, with no goal whatsoever, just random self-assimilation. To me it answers the ultimate question of free will. That’s really the origin of life.
Q: What is the biggest mistake or greatest lesson that you’ve learned?
Lipson: One lesson is: Don’t gamble your career on one idea. I think that in both academia and startups, there is this notion that you make a good decision and you put your eggs in that basket. In industry you focus on a product, and in academia you focus on a topic. Increasingly I don’t think that’s a good idea. I’ve learned that sometimes the best way to move forward is to do lots of things in parallel, because the reality is that the chance of any single expedition successfully working is small; maybe 10%.
If you try things one by one, perhaps after the third failure, you’re going to quit. But if you do 10 things in parallel, the chances are more likely than not, that at least one of them is going to succeed and that gives you energy to keep exploring. It’s counter to what you hear in academia and in industry, which is all about focus. But I say focus on lots of different ideas in a shallow quick way and drop the things that don’t work, expand the directions that do work and just sail west!