Mechanical Minds: Exploring the World of AI Robots

Speaker 1 [00:00:02] Hi, it’s John here. I’m in Vancouver, where over the years we’ve recorded some of our best episodes of Disruptors, looking at innovative technologies from quantum computing to artificial intelligence to hydrogen. Today, we’re going to explore the cutting edge of robotics and an exciting Canadian company called Sanctuary AI. Sanctuary was created in 2018 by Geordie Rose, Suzanne Gildert, Oliva Norton and Ajay Agrawal. They wanted to take on what’s perhaps the most complex tech challenge in front of humankind, which is how to build a human-like brain and system that’s capable of executing humanlike tasks in a safe way. Sanctuary is on a mission to create the world’s first humanlike intelligence in general purpose robots. Success to the team is when a person like you or me observes one of its Phoenix general purpose robots and is not able to tell whether a human pilot is directly operating the robot or whether it’s being controlled by an AI system. Sanctuary is backed by a range of companies and investors like Bell Canada and Magna, as well as Export Development Canada and Evok Innovations. It’s also at the forefront of a global race to apply artificial intelligence to robotics in a way that can make society better. This is Disruptors, an RBC podcast. I’m John Stackhouse. What you’re hearing in the background is the sound of engineers at sanctuary AI. They’re busy testing the Phoenix robot systems in one of their labs. Sanctuary AI is nestled in a quiet semi-industrial neighborhood of Vancouver. It’s kind of an unassuming office with lots of exciting things going on inside. I’m going to be talking with their co-founder and CTO, Suzanne Gildert.
Speaker 2 [00:02:13] Hi.
Speaker 1 [00:02:14] Suzanne. Great to see you.
Speaker 2 [00:02:15] Great to see you.
Speaker 1 [00:02:20] Great to have you on Disruptors.
Speaker 2 [00:02:22] Hello, thanks. I’m glad to be here.
Speaker 1 [00:02:24] So many questions to ask, but I want to start with a bit of your own story. And your accent is probably a giveaway, right? But tell us a bit about where you’re from and what brought you to Canada.
Speaker 2 [00:02:34] Yeah, so I grew up in the U.K. in a kind of medium sized town called Bolton, which is in the north of England. I guess science and physics is my first love. And then I also do a lot of art and stuff on the side as well. So I’ve got a bit of a left brain right brain thing going on there. I studied physics and electronics at university, did a PhD in quantum physics, and then at that point I decided to move to Canada to join a company called D-Wave. So that’s kind of where my foray into industry started, I guess.
Speaker 1 [00:03:06] We’ll get a bit into the D-Wave story. We had them on Disruptor a few years ago, but tell us about your initial interest in robotics.
Speaker 2 [00:03:14] Mm hmm. Well, I did hobby electronics with my dad, so I like building circuits and things like that. But when I first got interested in robotics properly was when I read a book by Jeff Hawkins called On Intelligence, and it was talking all about how the brain worked. And a lot of AI at that point was all about signals coming in, you know, through your eyes, through your ears. But that book made me think about the signals going out as well to your muscles to move your body. So I started thinking about AI is moving bodies in the world, and that usually means robots, because the robots are kind of the equivalent of bodies for A.I. brains.
Speaker 1 [00:03:52] When did you build your first robot?
Speaker 2 [00:03:54] I think it would probably be around 2012 or 2013 was the first proper one.
Speaker 1 [00:04:00] And since then it’s amazing that it’s been just a decade. But the transformation of robots is breathtaking.
Speaker 2 [00:04:06] Yeah, we’ve gone from having like tiny 3D printed sort of hobby robots all the way to one of the world’s most complex, full sized humanoid robots.
Speaker 1 [00:04:17] I’d say we all have visions of robots often given to us or imprinted on our brains by Hollywood. I’m going to age myself here by saying I grew up with a TV show called Lost in Space, where there was this classic robot character, which I still think of as my reference point in robots. Curious what in popular culture with robots inspired you?
Speaker 2 [00:04:38] I think the movie that inspired me the most was Chappie, which was kind of interesting, sort of like a gangster movie involving robots. But the thing I liked about the storyline there was that the way the AI system was developed to control the robot and the business and the industry and how that robot went out there into the world I thought was quite realistic, even though the movie itself was a bit crazy.
Speaker 1 [00:05:05] As you said, you came to Vancouver, you came to Canada for D-Wave, really interesting company in the quantum space. How do you go from quantum to robots?
Speaker 2 [00:05:15] So when I joined D-Wave, I was originally doing kind of physics experiments, but I found that I had a flair for actually programing the quantum computers and applying those quantum processes to problems. So, I got pretty good at quantum programing specifically for solving AI problems and then got really interested in AI. And it was around that same time that I read this on intelligence book, and that made me think about AI in terms of robots. So it was a bit of a weird pivot from quantum through AI to robots. And actually, now I’m getting back again interested in the quantum computing stuff because I think there are some ways we can apply quantum computing to robot and AI minds. So, it’s kind of coming full circle.
Speaker 1 [00:06:01] So before we get to Quantum meets A.I., these robots, tell us about Kindred, because that was a company you co-founded with Geordie Rose. Yeah, great entrepreneur and innovator who you’ve started sanctuary with. What was the vision for Kindred?
Speaker 2 [00:06:15] Hmm. Yeah. So the vision for Kindred was like, very similar to vision I’ve had all along for all these projects, which was we wanted to create a machine with humanlike intelligence. And we started by building these small humanoid robots. They looked kind of like Johnny five from Short Circuit. But the problem was because they were so small, they couldn’t really do anything. So what happened at Kindred was we took the control systems in the AI we’d been developing and we built a product robot that was very different from that. It was an e-commerce warehouse application. Where items came down a chute and then the robot had to sort the items into different bins. So we built a fleet of those and it was a very successful product, but we kind of lost a little bit of the original spirit of the mission of this sort of more humanlike A.I. So, although Kindred was very successful, we decided to spin out Sanctuary from that in 2018 to kind of go back to that early humanoid kind of vision.
Speaker 1 [00:07:21] Why do you want robots to be more human?
Speaker 2 [00:07:24] When I started doing this, this was a completely crazy idea. But if you look narrow, there’s actually a lot of humanoid prototypes being developed, and I think everyone’s now starting to come around to the same idea that if you want something that can do a lot of different general tasks in the world of work, a human like form factor is very useful because the world is designed for people. So what you find is if you build a humanlike robot, it just naturally fits into all of these environments that we already have that have been designed around us. And so I think the most general type of robot you could build is humanlike in form and function.
Speaker 1 [00:08:02] And do you envision robots also being human, like inside in terms of how they think and feel, not just how they walk and turn door handles?
Speaker 2 [00:08:12] Yes. Yes. That’s the goal of Sanctuary. It’s to build something that works very similar to how the human brain works and have that A.I. system controlling a robot that’s also very human like. And again, you might ask why? Like why not just create something that you can program and it does things like repetitive tasks. And the answer is, if you want this thing to be able to do lots of different kind of work and deal with all the different edge cases and strange things that can happen, you need something that’s very humanlike in its cognition. It has to be able to think like us reason about things. If something goes wrong, rethink, replan, deal with all the social beings that we interact with. So again, having something that had a human like mind is very useful for dealing with all these different situations.
Speaker 1 [00:09:00] How far are we from not only having robots like that, but interacting on a day-to-day basis with robots like that?
Speaker 2 [00:09:07] Yeah, it’s it’s a difficult question to answer because there’s a few key technology breakthroughs that I think have to happen before this will fully flourish. I mean, I personally believe by 2030 we’re going to definitely have these things walking around interacting with us on a day to day basis. Now, it could be sooner than that, but we can see a trajectory now for the breakthroughs that are happening to make this happen. And one of them is if you just look at the ecosystem, there is a lot more humanoids walking robots being developed. We’re seeing advancements in all these different areas of technology, from motors to simulation to computer graphics to A.I. algorithms. All these things are kind of a rising tide. So I think within a few years we’ll start to see the sparks of these things out there doing some work tasks. And then I think by the end of the decade we’ll start to see them being pretty general and behaving like us.
Speaker 1 [00:10:03] So hearing that this is possible by the end of this decade, I’m both excited and terrified. Which should I be?
Speaker 2 [00:10:10] We have the choice to steer where we want this to go. So one of the things I feel very strongly about is I want to build robots systems that do good in the world and get on well with people, interact well with those. They’re not in conflict with us. And in fact, one of the reasons that I want to make them like people is so that they understand us better and they’re able to communicate, live alongside us better rather than them just being some kind of alien machine. So a lot of people are scared of robots, but I think if they are behaving just like people are, then we will sort of get used to them very quickly. And I think they’ll become less and less scary the more people see of them.
Speaker 1 [00:10:59] Tell us what your robots can do today.
Speaker 2 [00:11:01] Right. So the systems we have today, we mostly operate them in what we call this teleoperation mode, which means as a person in control of the robot behind the scenes. So, a person wears some VR gear and they have a kind of an exo exosuit type rig that they sit in and they’re able to see what it sees through its camera. And so a lot of the tasks that we test, we put through this mode first because we want to make sure the robot can physically do the task. And then once we know it’s physically capable, look, we then move to trying to automate the task of trying to get the AI to then learn from that person to do the task. So we’re focused on fairly simple tasks, I’d say at the moment, like picking different objects from containers, placing them in other containers are on shelves or on fixtures. But if you look at most of what our brain is evolved to do is actually things like this majority of our brain is just all about understanding what objects are, how to use our hands to actually interact with the world and dexterously manipulate things and coordinate our entire body. We just take it for granted, but it’s actually quite difficult to get robots to do that.
Speaker 1 [00:12:13] So give us a sense of why that is so difficult, because robotic engineers have been at this for probably close to a century now, trying to master or develop a mastery of those basic mechanical skills, picking something up, moving it across the table, things we learn to do when we’re two years old and now we don’t even think about it. I’m picking up a glass of water here. I don’t even need to look at the glass to do that. And yet for a robot, that’s actually a complex challenge.
[00:12:43] Yeah, it’s very, very challenging. And it’s is kind of funny. You say like roboticists have been trying to do this for a century or something, but it actually took evolution a billion years to develop a brain that could do this. So I don’t think we’re doing too badly at the moment. But yeah, why is it hard? It’s hard because if you think about our sensors like our eyes and our ears, they’re kind of like cameras and microphones and there’s so much data coming into those sensors just, you know, megabits per second coming in. Our brain has to somehow turn that crazy stream of data into kind of a model of the world. And so that is just a really hard problem in itself. And then it has to do the opposite as well. It’s like has to turn a small piece of information into a stream of data that can control all the different motor signals. There’s just so many of them. So that’s why the problem is hard. It’s not impossible, but yeah, it just takes a long time.
Speaker 1 [00:13:42] I won’t take it for granted the next time I go for a glass of water.
Speaker 1 [00:14:00] This is fascinating. There must be many other companies also trying to do the same thing, solve the same problems. What do you believe is sanctuaries advantage in this so-called space race for robotics?
Speaker 2 [00:14:14] I think we’re kind of a systems integrator, so we take a lot of pieces of different technologies and we put them together in a whole system. So we don’t just build robots, but we build all these different modules in the brain that controls the robot. And we don’t want to reinvent the wheel. We don’t want to build a module if we can get something off the shelf that works. Large language models are a great example. So when large language models came along, we realized we can actually integrate that into the brains of our robot. So we kind of got a nice, prepackaged piece of technology that could enhance it already. So I think our advantage is we’re very flexible. Whenever something new comes out, we’re quickly able to integrate it into our full system.
Speaker 1 [00:14:59] So you’re essentially putting ChatGPT inside the robot’s brain.
Speaker 2 [00:15:03] Yeah, that’s one of the things we put in there.
Speaker 1 [00:15:05] What does that do to the robot?
Speaker 2 [00:15:08] Right, well, it allows it to, I don’t want to say understand language, but it gives it a mechanism for translating language into kind of actions are the things that our brain understands. So the programing language that our brain uses is not anything like natural language, or English or anything is quite mathematical. So what things like ChatGPT or large language models enable you to do is kind of translate between what we mean when we say things and how that mathematical brain understands things.
Speaker 1 [00:15:40] Eventually, from the way you describe it, is hard to imagine there being a difference between robots and humans. But is that going to be the case?
Speaker 2 [00:15:49] I mean, I think they’ll still always be differences. Early on in Sanctuary made some robots that actually looked much more like people, but now we’ve made them look a bit more like industrial droids.
Speaker 1 [00:15:59] Is that just to keep all of us from freaking out when one sees something that looks like us and acts like us?
[00:16:04] Yeah, there were a bunch of reasons. Mostly it was for functional purposes. Like it’s just easier to build something like that and kind of get it looking good. And but I mean, who knows, I think anything could happen in the future. I think we could have a mix of different types of robots. So there’ll be some that look very industrial. Maybe there’ll be some that end up looking more humanlike. Yeah, I can imagine a rich future of different robots.
Speaker 1 [00:16:30] I’m having visions of Westworld.
Speaker 2 [00:16:32] Right, right, right. I mean, the whole Westworld problem is, is when you have these sort of sci fi stories, they’re always very dystopian. People who write sci fi stories want to introduce a lot of conflict and drama into the storylines. But I caution everyone to think that they’re doing that for a reason to make a good story. It’s not reality.
Speaker 1 [00:16:53] Most of us just want a device that will take the garbage out or mow the lawn or shovel the driveway. Yeah. Well, let’s get into some of the applications. Where do you see the next generation of robots being actively used out in the economy or in society?
Speaker 2 [00:17:06] So I think what we’re targeting is places where there are a problems with labor shortages in areas where there just aren’t enough people to hire to actually do tasks. And then secondly, those tasks themselves are very repetitive, very dull. They can often be dangerous. So I think looking at these kind of tasks first is nice because you’re taking people out of harm’s way and then in many cases there are no people to do those jobs anyway. So the reason they haven’t been automated yet is that there are often edge cases where things can happen that a regular robot wouldn’t be able to deal with. So imagine someone has to pick a part out of a bin and then place it in a certain direction on another thing. Well, what happens if that part falls on the floor? A normal robot wouldn’t know what to do. But if you had a general purpose robot, it now would be able to deal with that edge case, like bend down and pick up the piece, put it back in the bin or whatever. So we’re looking at those kind of tasks first. But then eventually, I’d love to see these robots actually opening up entirely new job areas that we haven’t even seen before. So, initially we’d be looking at labour shortages in places where there aren’t enough people to do work. But then eventually I would like these systems to do jobs that don’t actually even exist yet.
Speaker 1 [00:18:25] A lot of the shortages that we all hear about are in areas that can be both repetitive but also have their own kinds of complexities. I was out with developers here in Vancouver listening to their challenges with getting construction workers and to know how many homes we need to build in this country over the coming decades. And yet, construction work is is very complex. It’s hard to move, move around to use the tools. Yeah. How far are we from having construction sites run by robots? And I can say the same about nursing wards and firefighting stations and the like.
Speaker 2 [00:19:02] It’s difficult to put exact dates on these things because there are some unsolved scientific problems here. But I’d say within a decade, I think we’ll definitely see those kind of more complex roles starting to be tackled by robots. There’s so much more productivity we could be kind of squeezing out of the economy if we had these labor units able to be put in places where there are gaps or shortages or even peak demand problems. So, I think it will be a much more flexible economy with a hybrid labor model where there’s like people working and then there’s robots working to fill in where all those gaps are.
Speaker 1 [00:19:48] How are we as a species going to need to adjust to work side by side with robots, maybe work for robots? I don’t know how it will evolve, but how do we need to change?
Speaker 2 [00:19:58] One of the reasons I want to make robots that are very human like people so we don’t have to change very much. I want them to be very familiar to us already. So I want to build systems where when you talk to a robot, it’s just like talking to a person, you don’t really need to modify your behaviour. Might be a little weird at first, but, you know, within a few days of working with that robot, you’re considering it like to be your buddy or a great coworker. If you think about when cell phones first came out, remember people thought they were awful and now everyone can’t live without their cell phone. Being right next to them is almost like a little pet or something, right? If it’s not there, you feel so distressed. So I think these robots will be like that. And it seems weird now from our vantage point, but I think we’ll get so used to working with them that will actually start to maybe even rely on them a bit or trust them.
Speaker 1 [00:20:49] Lots of companies come to you looking to develop use cases. Which sectors do you think are going to be most active in adopting robotics in the coming few years?
Speaker 2 [00:20:59] I think the biggest one, probably be manufacturing. There’s a few reasons for that. It’s at this sweet spot of these tasks that are very dull. People don’t like doing them, they can’t fill the roles and these tasks can often be quite dangerous as well. And the other cool thing about manufacturing is you can onshore because the cost will be lower. So we can essentially bring back a lot of manufacturing or start spinning up our own manufacturing here in Canada. We could become a leader in manufacturing, right. It’s sort of difficult to think about today because we think about certain countries as being where you get everything manufactured. But maybe with these new kind of robots, that doesn’t have to be the case. Maybe the game is going to change.
Speaker 1 [00:21:42] Absolutely, we trade with those countries often because of cheap labour. That’s our comparative advantage. And this is creating the ultimate in cheap labor. Although robots are made for free. How costly is it to produce a robot today?
Speaker 2 [00:21:57] It’s really a function of the volume of units you’re producing. So it’s kind of like cars. It can cost millions of dollars to create a new kind of like concept car. But once you’ve created it and it goes to mass manufacturing, you can produce something like, you know, 10 to 20,000 dollars or something. So the same is true of robots. They’re expensive right now. But I can see the cost of them coming down to similar to how cars, cost of cars scale over time. It’s really just a function of how many of them you’re building.
Speaker 1 [00:22:27] I don’t think many Canadians appreciate how advanced other countries are in robotics, particularly in in Asia. China has something like ten times the number of robots as the U.S. has and 100 times Canada’s rate. Where in the world excites you most in robotics right now?
Speaker 2 [00:22:45] So although lots of countries and now are able to create these robots, the real value is going to be in the companies that are developing the software. So I think here Canada has a huge advantage because we have a storied history in AI development, like the people who are known as the godfathers of AI did all that research here in Canada. So I think, you know, because we’re not classically known for manufacturing and hardware, I don’t think that’s really a problem because we can become a leader in the software and the control systems, which is where most of the value is.
Speaker 1 [00:23:18] So the hardware can come from somewhere else and they develop essentially the brains of the robot. We also have a tremendous advantage in our universities. Well, that’s why we have those godfathers. I think it’s six of the top 100 robotics schools in the world are in Canada, three of the top 50. That’s really impressive. How should companies or public sector organizations, for that matter, be thinking about robotics?
Speaker 2 [00:23:43] So I would have said a few years ago it was it was pretty niche and new. And the people that were interested back then were very early adopters. But what’s happened in the last few years is the category is actually now starting to emerge as a thing. So I’d say anyone who’s in business, especially when manufacturing or industrial type businesses to start having a look at this category because it’s now definitely on the horizon, it’s becoming a big thing. And I think the competitors will probably be looking at it as well.
Speaker 1 [00:24:20] The most successful companies that you get to work with on robotics, what do they do that differentiates them?
Speaker 2 [00:24:26] It’s a little early to know the answer to that question yet, because I still feel that the category is emerging. But I think what’s going to happen is the really innovative companies are going to be the ones that see the furthest out and don’t kind of get trapped in just, well, let’s build a special purpose robot for one task first and then we’ll expand from there. I think it’s going to be companies that are really looking at solving the fully general problem of how could you create a robot that could go around and do anything. So one thing that’s both very scientifically interesting, but also very easy to explain to someone is this idea of a domestic helper robot that can do all your chores for you. And it sounds kind of boring and mundane, but really trying to think about how to solve this problem. I think that’s where the companies of the future are going to get ahead, like taking a stab at those really hard problems that seem out of reach right now. Because in five years time when those companies are maturing, that won’t be out of reach anymore.
Speaker 1 [00:25:30] You’re very hopeful about robotics. I can feel it and hear it in your voice. What’s your greatest hope in the next decade? And equally, what’s your greatest fear?
Speaker 2 [00:25:40] My greatest hope for robotics is that we really do figure out and learn and grow these things, such as they get along with us and we’re able to work alongside them and not have that kind of conflict that you see in the sci fi movies. I think we need to have a mindset shift moving away from this thing that we fear to something that’s there to help us and really sort of like lift us to the next level as a civilization. My biggest fear is that we’re not able to be mature enough about this kind of technology and we overregulate, we start banning things and things suffer as a result. So I really do think with some scientific and AI advancements, we can address some of the problems that we’re having in the world, like supply chain issues, climate change, disease prevention and things like that. I think AI and robotics could really, really help with these areas where there are huge, open, unsolved scientific and geopolitical problems. And so my biggest fear is that we don’t embrace that and we kind of get stuck in a stalemate of everything being regulated and stopped.
Speaker 1 [00:26:52] As you’re building sanctuary. And by the way, I’d love to know why you call it Sanctuary. Are you needing to bring in different skill sets and ways of thinking to help you think through these fascinating but fairly deep challenges?
Speaker 2 [00:27:07] Yeah, absolutely. So on your first question, why do we call it Sanctuary? Well, it was designed to be a place where we could create this little bubble of unreality. So when we started Sanctuary, people didn’t believe these humanoid robots with humanlike minds could exist. So we wanted to create a place that was a sanctuary for people that actually believed in that and believed that these kind of systems can be built and used for good. So I think it’s great to have like a lot of scientists and engineers. But as we grow, it’s really important to understand how these systems will go out there in the world and integrate into business and industry and life in general. So, we need a lot of different people who have a lot of different expertise, you know, everything from hard core engineering to understanding creativity and design. And then people who really deeply understand customer needs and how these systems might be integrated. So yeah, lots of different skill sets I think are needed to really make it happen and bring it into the world.
Speaker 1 [00:28:08] We’ve talked a lot about robotics and it’s incredibly interesting, but I’d love to know a bit more your thoughts on humanity and how you hope these advancements in robotics can change and improve humans.
Speaker 2 [00:28:24] So the obvious answer is they can come into our everyday lives and really help. But I think the more kind of thought provoking answer to this question is we’re trying to build minds and brains that control robots. So in doing so, I like to think we’re actually understanding more about how our own minds work. And I think understanding ourselves and how we behave and act and how we process the world and process emotions and everything as people, I think we need more of that. I think we need more introspection. So ironically, by building A.I. minds, I think we’re actually going to understand ourselves better.
Speaker 1 [00:29:02] That’s a great message of hope to leave our listeners with, Suzanne. Thank you so much for being on Disruptors.
Speaker 2 [00:29:07] Well, thanks for having me. It was really fun.
Speaker 1 [00:29:13] Leaving the sanctuary labs, I’m back on the streets of Vancouver and it all feels very human again and yet very different. There’s people I think there are people driving cars and walking their dogs and going home from work and doing what humans do. That won’t change. But as we all know, technology is profoundly changing almost every aspect of our lives. We just need to look at the phones in our hands to realize that. And the power of robotics, especially when AI is applied, is almost beyond human comprehension. And that’s the exciting part when we apply our imagination to the power of technology, we’ll be able to tackle what today seem like insurmountable challenges. At least that’s always been the hope and promise of technology, and it’s really up to humans to take advantage of that. This is Disruptors. An RBC podcast. I’m John Stackhouse. Talk to you soon.
Speaker 3 [00:30:19] Disruptors, an RBC Podcast is created by the RBC Thought Leadership Group and does not constitute a recommendation for any organization, product or service. For more Disruptors content, visit our RBC.com/disruptors and leave us a five star rating. If you like our show.

This article is intended as general information only and is not to be relied upon as constituting legal, financial or other professional advice. A professional advisor should be consulted regarding your specific situation. Information presented is believed to be factual and up-to-date but we do not guarantee its accuracy and it should not be regarded as a complete analysis of the subjects discussed. All expressions of opinion reflect the judgment of the authors as of the date of publication and are subject to change. No endorsement of any third parties or their advice, opinions, information, products or services is expressly given or implied by Royal Bank of Canada or any of its affiliates.