Siddhartha “Sidd” Srinivasa has spent his career at the intersection of robotics research and real-world deployment. He helped start Berkshire Grey, launched Amazon’s Robotics AI group, helped re-architect Cruise’s autonomy stack, and leads the Personal Robotics Lab at the University of Washington as a professor.
Now he’s adding a new role: venture partner at Madrona Venture Group, where he’ll help shape investment themes and evaluate startups in robotics and embodied AI at the Seattle-based firm.
GeekWire caught up with Srinivasa for a conversation about the state of robotics, opportunities for entrepreneurs, and why he believes Seattle can become a global leader in AI and robotics. The interview was edited for brevity and clarity.
GeekWire: Why did you decide to join Madrona?
Sidd Srinivasa: “I want to give back to the Pacific Northwest and make the Pacific Northwest an even bigger superpower in AI and robotics. I’ve been incredibly fortunate to work with some really wonderful VCs during my time at Berkshire Grey, and it felt like time for me to both give back as well as learn what it takes to be a venture capitalist — to think about how to evaluate companies, to develop a sound thesis on where AI and robotics should go next.”
Where is robotics right now in terms of progress?
Srinivasa: “Hardware has become commoditized. A humanoid robot used to cost a couple of million dollars, only accessible to top labs. Now you can get one for $15,000 to $20,000. A large part of that is because electric motors have gotten better and cheaper, and our ability to produce them and to be able to scale them has gotten much better.
It’s also the compute. I have more compute on my phone now than I did on the biggest computer I could find during my Ph.D. That’s pretty significant.
Building large foundation models is also commoditized. And what these foundation models help us do is deal with all the common sense stuff that goes into getting any task done.”
What about the current challenges?
Srinivasa: “The place where we have a huge problem with robotics right now is that though we are able to address a lot of the common sense tasks like folding your laundry or clearing your table — those are not the important tasks people would pay money for. This is a fundamental issue that a lot of robotics companies are facing right now. Yes, we have pretty amazing demos, we have cool videos of humanoids picking up an apple and putting it down. But who wants to pay for that? Not me.
Robotics has a last-mile problem. We can go from zero to one, one to 10, 10 to 50 — but very few times have we actually been able to go the last mile. We certainly did that at Berkshire Grey. We certainly did that at Amazon and with fulfillment. But beyond fulfillment, I haven’t seen any great success in robotics.”
What’s missing technologically?
Srinivasa: “We need breakthroughs in generalization. The other day, when I was staining my fence, by the time I got 20% done, I was an expert stainer of fences. That adaptiveness is what we need robots to do efficiently. Reinforcement learning is a way to do that but it is very data inefficient. So being able to truly generalize novel scenarios and being able to learn on the fly while being efficient and safe — safety is really important.
I thought we would solve self-driving before we solve general purpose robotics. And we haven’t solved self-driving yet. We’re close, but not there yet. And I would say general purpose robotics is certainly five years behind self-driving in terms of capability and technology.”
What about humanoids?
Srinivasa: “I am not sure whether the humanoid form factor is the right form factor to align around. Humans have two arms and two legs just ’cause — there’s nothing theoretically optimal about us. Maybe what you need is a robot with three arms, or a robot with seven fingers. I do believe that one day we will have capable robots performing useful tasks in our home. Whether that’ll be a humanoid or a quadruped or a three-hand, three-arm robot — I don’t know.”
Which sectors look closest to achieving viable unit economics?
Srinivasa: “There’s still a lot of juice left to squeeze in fulfillment. There’s over 50-to-60% of fulfillment that is yet to be automated — things like packing an Amazon box full of items. We need to be able to deal with that kind of high density. And I think last mile fulfillment is a space that is really ready to be disrupted.
I’ve been really interested in computational agriculture from an observability point of view — knowing what’s going on in a greenhouse without having to walk the line. And commissary kitchens that make lunches that are delivered or airline meals. If we are able to build robots that can be very easily repurposed and retrained to perform multiple tasks and have a small form factor, then I think that’s a space that can be really disrupted.”
What advice do you have for robotics founders?
Srinivasa: “There has to be a pull for robotics, not a push. If you’re pedaling robotics to people who reluctantly accept it, it’s never going to work. The pull has to come from a real customer need.
The second thing I would say is that there is this tendency to believe that robots can follow the software model. Maintaining, servicing and taking care of software is really easy. But that’s not the case with real, physical robots. I would think very carefully about not just the actual application, but also the modality by which you are putting your robots out there.”
How do you think about robotics and job displacement?
Srinivasa: “Ever since I started working on robots, I’ve been very rightfully asked: will robots take our jobs? The lazy argument is to say the Industrial Revolution created more jobs than it took away, so robotics will, too. But it’s not the people who lost their jobs who got the new jobs. Pittsburgh is a great example — steel workers didn’t become tech workers. So we have an entire generation of people who lost their jobs, who we did not re-skill to be able to take on the new roles that were coming up. We need to be committed to workforce re-skilling such that we can enable and empower the workforce to be able to work with and around new technology that comes about. It’s not just a nice thing to do. It’s a societal imperative.”
Any thoughts on Seattle as a robotics hub?
Srinivasa: “I think one of the things that’s incredibly appealing about Seattle, and particularly about the University of Washington, which I think has been a pioneer in this, is just how closely faculty are able to work with industry while still maintaining and honoring their faculty position.
I love Seattle. I think it’s also incredibly valuable for dual careers. And I think it’s a great place to start a company, not just because of Madrona, but also because of so many other VCs, as well as the huge amount of talent that’s out here. So I think it’s time to make Seattle the best in the world in AI and robotics.”
