In this episode Byron and Dr. Rosenberg talk about AI and swarm intelligence.
- Subscribe to Voices in AI
- iTunes
- Google Play
- Spotify
- Stitcher
- RSS
Guest
Dr. Louis Rosenberg is the CEO of Unanimous AI. He also holds a B.S., M.S., and a PhD in Engineering from Stanford.
Transcript
Byron Reese: This is Voices in AI, brought to you by GigaOm. I’m Byron Reese and today I’m excited that our guest is Louis Rosenberg. He is the CEO at Unanimous A.I. He holds a B.S. in Engineering, an M.S. in Engineering and a PhD in Engineering all from Stanford. Welcome to the show, Louis.
Dr. Louis Rosenberg: Yeah, thanks for having me.
So tell me a little bit about why do you have a company? Why are you CEO of a company called Unanimous A.I.? What is the unanimous aspect of it?
Sure. So, what we do at Unanimous A.I. is we use artificial intelligence to amplify the intelligence of groups rather than using A.I. to replace people. And so instead of replacing human intelligence, we are amplifying human intelligence by connecting people together using A.I. algorithms. So in laymen’s terms, you would say we build hive minds. In scientific terms, we would say we build artificial swarm intelligence by connecting people together into systems.
What is swarm intelligence?
So swarm intelligence is a biological phenomenon that people have been studying, or biologists have been studying, since the 1950s. And it is basically the reason why birds flock and fish school and bees swarm—they are smarter together than they would be on their own. And the way they become smarter together is not the way people do it. They don’t take calls, they don’t conduct surveys, there’s no SurveyMonkey in nature. The way that groups of organisms get smarter together is by forming systems, real-time systems with feedback loops so that they can essentially think together as an emergent intelligence that is smarter as a uniform system than the individual participants would be on their own. And so the way I like to think of an artificial swarm intelligence or a hive mind is as a brain of brains. And that’s essentially what we focus on at Unanimous A.I., is figuring out how to do that among people, even though nature has figured out how to do that among birds and bees and fish, and have demonstrated over millions of years and hundreds of millions of years, how powerful it can be.
So before we talk about artificial swarm intelligence, let’s just spend a little time really trying to understand what it is that the animals are doing. So the thesis is, your average ant isn’t very smart and even the smartest ant isn’t very smart and yet collectively they exhibit behavior that’s quite intelligent. They can do all kinds of things and forage and do this and that, and build a home and protect themselves from a flood and all of that. So how does that happen?
Yeah, so it’s an amazing process and its worth taking one little step back and just asking ourselves, how do we define the term intelligence? And then we can talk about how we can build a swarm intelligence. And so, in my mind, the word intelligence could be defined as a system that takes in noisy input about the world and it processes that input and it uses it to make decisions, to have opinions, to solve problems and, ideally, it does it creatively and by learning over time. And so if that’s intelligence, then there’s lots of ways we can think about building an artificial intelligence, which I would say is basically creating a system that involves technology that does some or all of these systems, takes in noisy input and uses it to make decisions, have opinions, solve problems, and does it creatively and learning over time.
Now, in nature, there’s really been two paths by which nature has figured out how to do these things, how to create intelligence. One path is the path we’re very, very familiar with, which is by building up systems of neurons. And so, over hundreds of millions and billions of years, nature figured out that if you build these systems of neurons, which we call brains, you can take in information about the world and you can use it to make decisions and have opinions and solve problems and do it creatively and learn over time. But what nature has also shown is that in many organisms—particularly social organisms—once they’ve built that brain and they have an individual organism that can do this on their own, many social organisms then evolve the ability to connect the brains together into systems. So if a brain is a network of neurons where intelligence emerges, a swarm in nature is a network of brains that are connected deeply enough that a superintelligence emerges. And by superintelligence, we mean that the brain of brains is smarter together than those individual brains would be on their own. And as you described, it happens in ants, it happens in bees, it happens in birds, and fish.
And let me talk about bees because that happens to be the type of swarm intelligence that’s been studied the longest in nature. And so, if you think about the evolution of bees, they first developed their individual brains, which allowed them to process information, but at some point their brains could not get any larger, presumably because they fly, and so bees fly around, their brains are very tiny to be able to allow them to do that. In fact, a honeybee has a brain that has less than a million neurons in it, and it’s smaller than a grain of sand. And I know a million neurons sounds like a lot, but a human has 85 billion neurons. So however smart you are, divide that by 85,000 and that’s a honeybee. So a single honeybee, very, very simple organism and yet they have very difficult problems that they need to solve, just like humans have difficult problems.
And so the type of problem that is actually studied the most in honeybees is picking a new home to move into. And by new home, I mean, you have a colony of 10,000 bees and every year they need to find a new home because they’ve outgrown their previous home and that home could be a hole in a hollow log, it could be a hole at the side of a building, it could be a hole—if you’re unlucky—in your garage, which happened to me. And so a swarm of bees is going to need to find a new home to move into. And, again, it sounds like a pretty simple decision, but actually it’s a life-or-death decision for honeybees. And so for the evolution of bees, the better decision that they can make when picking a new home, the better the survival of their species. And so, to solve this problem, what colonies of honeybees do is they form a hive mind or a swarm intelligence and the first step is that they need to collect information about their world. And so they send out hundreds of scout bees out into the world to search 30 square miles to find potential sites, candidate sites that they can move into. So that’s data collection. And so they’re out there sending hundreds of bees out into the world searching for different potential homes, then they bring that information back to the colony and now they have the difficult part of it: they need to make a decision, they need to pick the best possible site of dozens of possible sites that they have discovered. Now, again, this sounds simple but honeybees are very discriminating house-hunters. They need to find a new home that satisfies a whole bunch of competing constraints. That new home has to be large enough to store the honey they need for the winter. It needs to be ventilated well enough so they can keep it cool in the summer. It needs to be insulated well enough so it can stay warm in cold nights. It needs to be protected from the rain, but also near good sources of water. And also, of course, it needs to be well-located, near good sources of pollen.
And so it’s a complex multi-variable problem. This is a problem that a single honeybee with a brain smaller than a grain of sand could not possibly solve. In fact, a human that was looking at that data would find it very difficult to use a human brain to find the best possible solution to this multi-variable optimization problem. Or a human that is faced with a similar human challenge, like finding the perfect location for a new factory or the perfect features of a new product or the perfect location to put a new store, would be very difficult to find a perfect solution. And yet, rigorous studies by biologists have shown that honeybees pick the best solution from all the available options about 80% of the time. And when they don’t pick the best possible solution, they pick the next best possible solution. And so it’s remarkable. By working together as a swarm intelligence, they are enabling themselves to make a decision that is optimized in a way that a human brain, which is 85,000 times more powerful, would struggle to do.
And so how do they do this? Well, they form a real-time system where they can process the data together and converge together on the optimal solution. Now, they’re honeybees, so how do they process the data? Well, nature came up with an amazing way. They do it by vibrating their bodies. And so biologists call this a “waggle dance” because to humans, when people first starting looking into hives, they saw these bees doing something that looked like they were dancing because they were vibrating their bodies. It looked like they were dancing but really they were generating these vibrations, these signals that represent their support for their various home sites that were under consideration. By having hundreds and hundreds of bees vibrating their bodies at the same time, they’re basically engaging in this multi-directional tug of war. They’re pushing and pulling on a decision, exploring all the different options until they converge together in real time on the one solution that they can best agree upon and it’s almost always the optimal solution. And when it’s not the optimal solution, it’s the next best solution. So basically they’re forming this real-time system, this brain of brains that can converge together on an optimal solution and can solve problems that they couldn’t do on their own. And so that’s the most well-known example of what a swarm intelligence is and we see it in honeybees, but we also see the same process happening in flocks of birds, in schools of fish, which allow them to be smarter together than alone.
So let’s take that apart a little bit. And, for the record, I raised honeybees.
Oh, wow.
And so I can sing their virtues and all these clever things that they do. You know, they make this big beard outside the hive when it’s hot; they all flap their wings and circulate air in. They can communicate to each other where pollen is relative to, like, where the sun is in the sky and all these things. And so, do you believe we understand the mechanism? Like you can observe it and say, “The bees go out, they find a new home, they have the old queen, they all fly out, they all come back, they dance and then they pick the right one.” So is it that we know how that happens? Do we know how that happens?
We do. And so biologists have done some amazing research—a lot of it’s been done at Cornell University—and what they literally do is they will watch colonies of honeybees where they’ll paint little dots on the bees of different colors so they can actually keep track of which bee is which. And this type of research actually started in the 1950s, when people were literally watching bees that had been painted different colors to see how they go out, what potential sites they visit, then they come back to the colony and they vibrate their bodies and the magnitude of their vibrations indicate how strong they support the different options that they’ve considered and then other bees get influenced. And, like I said, this first started by people watching bees by hand.
In more recent years, they’ve used vision systems and computers to watch the bees, but biologists have done really remarkable research to understand in great detail how these decisions get made. And they get made as a system with feedback loops, where they’re basically negotiating in real time, vibrating their bodies and different factions form. And so a faction of bees will form vibrating in support of one option. Well, another faction will form vibrating in support of another option. And there’ll be lots of different options with factions, basically, growing and shrinking until they can converge on that one solution that maximizes their collective support. And that solution is almost always the best possible solution.
Let me ask the question this way. One biological interpretation of that, an evolutionary one would be, through some freak chance occurrence, one hive somewhere started doing this thing and it worked. Every other hive died out or co-mingled with this one and they just lucked into… to say they’re not really intelligent because they can’t solve any other problem. You know, they can’t—I mean, they can solve different kinds of problems, but it isn’t diversifiable intelligence.
You know, I remember, there was this Coke machine at school when I was young and it was really erratic but if you put a quarter in and then kind of hit the button a few times and then put your other quarter in and hit the button a few times, it would work better. Well, we just learned that trick. We didn’t know anything about—we weren’t intelligent, we just happened to stumble across the trick and the trick spread around and that was it. So I guess my question is, do you think that all of those bees are analogous to neurons in your brain? No one neuron knows anything. Collectively, though, they create a consciousness with all these attributes like a sense of humor and all of that. So that’s one thing. And so I would ask you, is that collective consciousness aware? I mean, is it something that is emergent and is aware of the world and is making a decision, or is it just some weird fluke of behavior and they aren’t really thinking anything?
Yeah, that’s a great question and it’s something I think about a lot. Mostly from the perspective that to us humans, the behavior of bees is pretty foreign, and so it’s hard for us to even notice, let alone appreciate, the sophistication of how bees are behaving in their world. And so we do know that they solve this finding a new home problem by working together as a swarm intelligence.
There are actually lots of other problems they solve as well in terms of finding optimal places to direct their scout bees for finding pollen or when a whole swarm has to fly together to find a new location. So there’s lot of things in a bee’s world that they need to do intelligently. To us humans, it’s easy for us to think, that’s not very exciting intelligence, because we interact in a pretty different world than bees do. The same thing happens when you look at ants or fish or birds. You know, they’re solving a range of complex problems by thinking together in systems, but it’s hard for us humans to think about that as something sophisticated. And we could say, “they’re just solving this one simple problem. It’s a trick or it’s—” I don’t believe it is that way. I believe that when they are forming this system, they are creating an emergent intelligence that has its own set of behaviors. And if we were bees, we would think, “Oh, that colony has a certain personality, it behaves a certain way, and that colony’s behavior is different than another colony’s behavior.” To us, all colonies seem the same, but I’m sure to an alien, all humans would seem pretty similar. The subtle differences in our personality are very apparent to us, but to an outsider, we humans might look like we’re very predictable in how we solve problems or how we react to things. I think the same thing is apparent in schools of fish or in colonies of ants. I believe that each colony or each school or each flock is its own emergent intelligence with its own personality and its own unique ability to solve problems and it’s very difficult for humans to appreciate that, because we are not in tune with the subtleties of their world the way that they are.
So, to be clear, you say your feeling is that, you suppose or your best guess is that they really are, to one degree or another, conscious. And if the problem changed, if there was a law that was passed in the whole world that all bee colonies more than eight feet off the ground have to be destroyed, that you believe the bees would learn that. That over and over, all these hives, I mean, all these hive swarms and any of them would go high up in, like, a tree, on a big branch, they’d get destroyed, anything that’s low… That that system isn’t some static, rigid thing but that there’s a consciousness in that that would say, like literally say, “Oh, the problem has changed. No longer are high places any good.”
Right. So that’s a great question, and part of that comes to what is the time constant for them being able to discover that. And so that’s the type of—
My example was really not—
Right, well—
But the basic idea is, can that individual hive learn new behavior if the old behavior doesn’t work? Or do you think it’s hard-coded in being a bee, it’s just their DNA?
Yeah. It’s a really good question, and it’d be really interesting to see if there are examples that have specifically been studied by biologists. I think for honeybees, at least, the example of finding a new home, they get one shot at it, right, each season and if it doesn’t work out, they probably end up dying as a colony.
Right.
And so it is a behavior that would get learned over time, through the evolution of their species. It makes me wonder if there are examples that have a shorter time constant where either, bees would learn as a system that the pollen that they get from certain types of cultivated fields maybe isn’t as rich as other kinds of natural settings, and do they learn over time to direct their colony towards the plants that have better pollen sources? I would suspect that they do, do that. So what they’re able to do as a system is compare options and discover which are working best for them and then learn over time to guide in that direction. It’s really interesting and I would bet that there probably are some examples that biologists have discovered. You know, it’s very difficult research for biologists to do because we—
They get stung a lot.
Right, they get stung a lot.
You know, there’s an old legend that if a beekeeper dies, somebody needs to go out and tell the hives, to tell each of them that the beekeeper is dead.
Is that a true legend or you’re making that—?
It’s true. No, I’m very serious. Like I said, I raised bees.
Yeah.
I studied lore and there’s a lot of stuff like that. You know, there are a lot of things like that, that soon the bees are more than the sum of their parts. So let’s talk briefly about emergence. So I have a new book that comes out in like a week.
Wow.
And it’s about consciousness and where it comes about, how it comes about in humans. And, you know, everybody knows what consciousness is: it’s the subjective experience of being yourself. It’s warmth, it’s how you feel temperature and all of that. But there isn’t a broad agreement on how it comes about. Nobody really knows why a bunch of neurons are able to perceive the world as opposed to just exist in it. You know, it’s a quantum phenomenon. It’s a product of emergence. Consciousness is imbued in everything. It’s a fundamental force of nature. I found like eight different ways that you can explain it. And presumably you don’t have to know how the bees do it; you just kind of want to copy their trick. But do you have a sense as to how they do it?
You mentioned emergence. So there’s two kinds of emergence. There’s weak emergence—where once you see what happens, you can kind of figure it all out, but there’s no break in the laws of the universe that gets you this result—and then there’s strong emergence. And some people think consciousness is the only kind of strong emergence. And it says no matter what you know, you cannot ever draw a line between how a group of cells eventually becomes conscious. So if it’s emergent, do you think it’s a strong emergence? And I only want to talk about this a little bit more because the meat of our conversation is: Now how do you take all of that and apply it to our world.
Right. And so these are great questions and from my perspective, as you said, nobody really knows the how, but we do know the what. And one of the what’s that we know is that a group of neurons, which are very, very simple on their own, and it’s hard to even consider a single neuron to be intelligent on its own, but it does have processing capability. If you take a large enough number of simple neurons and you connect them into a sophisticated system, and a system where they’re very deeply connected—and so, in a human brain, we have about 85 billion of these neurons and each one is connected to thousands of others, so there’s over 100 trillion connections, so massively connected. We do know that at that level of connectivity, human intelligence emerges. We know because we are the proof of that, but we have no idea exactly how or why that happens.
We also know that much smaller groups of neurons can have intelligence emerge in other organisms, organisms that don’t have 85 billion neurons, but even a single honeybee. A single honeybee is intelligent, I would presume it’s conscious, I would presume it’s self-aware, and it has only 1 million neurons. And so we do know that this happens. We also know that if we have—one of the fundamental principles of these groups of neurons is that they are systems. They are real-time systems with feedback loops where changes in one neuron will propagate through the system and affect lots of other neurons and even feedback to the original neurons, that you have these systems with feedback loops. And when we model those systems with feedback loops, we actually find that swarms follow very similar principles.
It’s following those similar principles, just up a level, where instead of connecting groups of neurons together into systems, you have groups of brains that are being connected together into systems with feedback loops where changes in a single organism can propagate through the system and affect many other organisms and even propagate back to the original organism. And so, again, it’s a very similar pattern—it’s just up one layer. And my presumption is that if you have enough organisms and they are connected in a rich enough system, they will be able to behave as a unified entity, they’ll behave as a single-thinking machine, and it will have its own unique personality based on the make-up of the individuals who are participating, and it will have its own form of intelligence.
I see that process of emergence where intelligence is coming out of a system at the swarm level, at the hive mind level, really being analogous to intelligence emerging at the neuron level. But we have much less experience in understanding emergence of intelligence at the hive mind level because the biological examples are so foreign to us, like bees and ants and fish. That said, at least from my experience, as we start connecting groups of people together, we start to see that these groups of people become smarter together than they would be alone, and they start to behave with a unique personality that’s different, that you couldn’t assign to any one of the individuals. It’s its own unique group personality.
Well, think about a riot, you know, where everybody just goes berserk and starts smashing everything up. Like, their ball team won that night and so riots start. Is that humans acting as a unit, they’re all influencing each other about what they do, taking on different personality aspects and doing some goal that no one of them could have done—smash up the city?
Yeah. So, you know, in some sense you’re referring to a mob, right?
Right, right.
And a mob is an example of a group that has an emergent behavior in it, and that emergent behavior is negative and that emergent behavior has a personality that’s different than the individuals, how they would behave on their own. So a mob is a good example of an emergent property of a group that’s interacting. The principles of a mob actually follow more a different natural example of a herd. And so, in nature, there are herds and there are swarms. And a herd is a group where there tends to be leaders and followers. The thing about a herd is that, if you have, say, a herd of sheep, and you spook one sheep and it takes off running, all the other sheep will sense that and take off running as well, and you have a single leader and a whole bunch of followers. And the problem with a herd is that, if the single leader that gets spooked jumps off a cliff, the rest of the sheep could jump off a cliff and there actually are documented examples of that happening in nature.
And so herding or herd mentalities, which tend to emerge in prey animals because it’s a defense mechanism, is actually more similar to this mob mentality, where there’s this emergence of a behavior that kind of spirals out of control. Swarms in nature tend to behave differently, where in a honeybee swarm there are no leaders or followers and behaviors don’t propagate through in the same way. All of the participants are interacting at the exact same time and so it’s more of a thoughtful deliberation than a kneejerk reaction.
So in nature it’s interesting to see these different behaviors appear in other organisms. You know, in a prey organism where a herd could make bad decisions but what’s more important is it can make a kneejerk reaction and escape because they’re prey, you could end up with a mob mentality emerging. In a swarm, like bee swarms, where what’s more important is thoughtful deliberation, the process is parallel rather than serial and you end up with a group that can make, I would say, an enlightened decision that’s better than the individuals would have made because they have the luxury of time to make that decision.
Well, but one would not assume that even though the mob has this emergent personality, for lack of a better word, nobody that I know of would argue that a mob has an emergent consciousness as well. Like a mob knows every bit of what it’s doing and it has a will and it has a direction and it has an intention. Is that the case? And second question: are there naturally occurring swarm behaviors that humans exhibit?
Right, so two really interesting questions. I mean, it’s a great question of, “does a mob have a consciousness?” And I would say that it’s not self-aware in the way we think of ourselves as being self-aware. Partly because it doesn’t have the feedback loops as a system. I mean, it has localized feedback loops. If you’re in a mob, you’re becoming influenced by that mob and you’re perceiving what’s happening, and it’s probably changing your individual sense of identity, and so it starts to get into the philosophical idea of, what is consciousness?
But, you know, as we’ve all experienced being in groups, I think it’s difficult to assign the sense of self to this larger organism. That said, if you were outside of the mob, you know, you’re assigned with trying to control this mob, right? You’re trying to corral this mob, if you could see its behavior from above, you would start to sense its personality, you would start to sense how it would react. You know, you would try one thing and it would react a certain way and it would start to have these predictable behaviors. And if you could have this kind of God’s eye view of this mob, you would probably start to perceive it as what biologists would call “a super-organism,” because it’s behaving as a unit and it’s behaving as a unit in predictable ways based on how you interact with this mob. You know, if you were putting up barriers, this mob would behave in a certain way with respect to those barriers. And this mob could probably learn over time. And so, from the outside, it could be viewed as a super-organism that reacts in certain predictable ways and then learns to start reacting in less predictable ways and each individual in that mob would probably have its sense of identity changed as it’s part of that system. But the one piece that gets philosophical is if we say, does that system have its own sense of self? And we have no reason to believe that it does, although it is behaving as a super-organism.
And do naturally occurring swarm behaviors happen in day-to-day life?
And so the thing about human groups is that we definitely are forming systems, but they’re not the same type of closed-loop systems that birds and bees and fish do. Basically, people can’t waggle dance but we do work in groups. And so the types of swarm behaviors that we would see in bees, we don’t see naturally occurring in people. If we look at other types of behaviors, there’s classic behaviors called “flocking,” which goes more to navigation, right? You know, schools of fish and flocks of birds can make decisions as a group, which we would call “swarming” but they also can navigate as a group, and we would call that flocking. And people can do that. You know, if you watch crowds of people, they can navigate as a group and people are well-adapted to walking in a crowd and following the people in front of you. And you might not even know where the group is going but you can behave as part of that type of group. And we see it in human behaviors because people evolved from other organisms that have strong tribal behaviors or even herding behaviors. And so, I think we do see group behaviors that are similar to what we see in other organisms, but when we’re thinking about the decision-making abilities of groups like honeybees, we don’t see it in nature, mostly because people can’t form those really tight feedback loops by waggle dancing where large groups can all interact at the exact same time.
Okay. Well, now that we’ve established that, if the hive can behave a million times smarter than one member, how do we build the superintelligence of humans that’s a million times what any of us can do? How do you actually do that?
Yeah. So that’s a great open question and something that we’ve been researching for a number of years and, again, we always look to nature for inspiration, and we say, well, the one thing we know that natural systems do is they form these real-time systems with feedback loops where they can all behave at the exact same time and converge together on solutions. And so what we’ve been doing at Unanimous A.I. is enabling groups of people over the Internet, who can be anywhere in the world, enabling them to interact as a unified system to answer questions.
And once we enable groups of people to work as a unified system to answer questions, we can start to understand… Do they get smarter? And how much smarter do they get? And what we’ve found is that even relatively small groups, when they’re enabled to work together as a system in real time, can become statistically significantly smarter. Now, they’re not a million times smarter, they’re not a thousand times smarter, but they are significantly smarter. So I can give you an example. One of the things we do at Unanimous A.I. is we have groups of people make predictions and forecasts as a swarm. And we do that not because that’s the only measure of intelligence, but it’s a verifiable measure of intelligence because we can do a controlled experiment. We can have a group of people make a prediction on their own, and we can allow that same group of people to make predictions by converging together as a system.
And so a study that we recently did with Oxford University was to look at predicting soccer games in the English Premier League. And so we looked at groups of sports fans to say, can they predict—? We had them predict 50 soccer games over a period of 5 weeks—so, you know, a really good statistical sample—and we had that group of people predict first as individuals, so they bring these 50 games over 5 weeks by just giving us their individual predictions. And they also did it by working together as a swarm, where they would converge together basically pushing and pulling on each other, using a model that is very similar to honeybee swarms.
What we found was that these individuals, when they predicted on their own, were 55% accurate in predicting the outcome of the games as individuals. And 55% doesn’t sound great, but in English soccer—or football, as they would call it—the games can end in a win, a loss or a tie. And since, because they allow ties, 55% accurate in predicting the outcome of the game is actually pretty good. And so the individuals were 55% accurate on their own. The same exact people, when they were thinking together as a swarm, jumped all the way up to 72% accurate in predicting these games, so that was 131% amplification of intelligence with a swarm of people that was only about 50 individuals.
So 50 individuals forming a swarm, using the current technology that we have, saw 131% amplification of intelligence. That said, we look at this technology space as being in its infancy. And if very simple systems can achieve that level of amplification, and we look at how much amplification honeybees achieve—which we think is massive considering how they can solve very complicated problems even though they have extremely simple brains—it’s our presumption that this technology space will enable larger and larger human swarms, and will enable better and better interactions among the members, and will be able to go from 131% amplification of intelligence to 200% to 500%, to thousands of times of amplification of intelligence. It’s a matter of developing more sophisticated systems, and larger systems, and systems where the interaction among members is more sophisticated.
Well, the way you describe it with, like, everybody kind of pushing and pulling, it sounds like a Ouija board. Like everybody’s hands on the marker and the question comes out and the marker is kind of influenced to get all these different places. I’m sure it isn’t a Ouija board, but what is a human waggle dance? What are people doing when they are coming together and waggle dancing and convincing each other or not?
Right, so that’s a great question. And so the systems that we do build today do have an aesthetic that makes people think of a Ouija board because you have a set of options that people are considering among, and we have an interface that’s very intuitive to allow groups of people to guide the system towards various options. And when bees make a decision among different potential home sites, they have a similar scenario, where they’re trying to decide among a large set of options. And what biologists have found is that bees can express their opinion by indicating which option they prefer most and also indicating their level of preference. And when these systems converge, they’re looking at how the bees are expressing their direction and magnitude of their preference.
And so what we created for humans is to allow people to do the same thing over the Internet. And because we wanted to allow a very natural interface that anybody could interact with, we figured, well, the best way to do that is to use a standard mouse or touch screen. And so we currently allow people to control a little magnet on their screen that conveys both the direction and magnitude that they want the swarm to go. And so you can have a hundred people, each controlling a little magnet that’s basically pulling on the swarm and they’re indicating by the position and orientation of their little magnet which direction they want the swarm to go. But when the swarm starts going in a direction and they realize it’s not going to go to their favorite choice, people start switching and what happens is that you have a group of people who are basically pulling to some options and then switching to other options and we have A.I. algorithms that are basically watching their behaviors and determining their levels of conviction. And so we’re combining all of their sentiments in a way that accounts for differences in their levels of confidence or differences in their levels of conviction and guides the swarm based on how they are expressing themselves. But because it’s a feedback loop, everybody’s adjusting to everybody else. And it’s a little bit hard to describe in words.
On our website, we have lots of video examples of what these swarms looks like, which is just www.unanimous.ai, but the key for us was to enable groups of people to interact in an intuitive way to express their intent and react to everybody else in real time. And once they have those basic abilities, they’re able to converge on answers to problems that maximize their collective confidence and their collective conviction in ways that allow them to, basically, have the best combination of their knowledge and wisdom and insight and intuition. And, again, we think this technology is in its infancy, and so while we’re currently enabling groups of people to do this using standard computers with a mouse or a touch screen, over time the way groups of people will interact in these types of systems will include voice and face recognition and, ultimately, brain interfaces because the better we enable people to express themselves and interact as a system, the more powerful these swarm intelligence hive minds will become.
What is your use case right now? As a company, are you making these tools and other people use them to solve whatever problems they want? So are you a platform provider? Are you a product provider? Or you’re going to actually use it yourself and, you know, just pick the perfecta each time at the horserace or whatever?
Right. So we are a platform provider, but right now the customers we have are large corporations that want to generate intelligence from groups and they want to, basically, amplify the intelligence of groups to answer questions that are relevant to their business. And so we basically have a service we call Swarm Insight where we generate insights for large companies by basically bringing together swarm intelligence on particular topics. And so our current customers are large companies like CNN, Deloitte, eBay, Red Bull, Boeing. We’ve done projects for the U.S. Navy. And so the type of project, if you’re a company that makes a product and you want to understand how consumers will react to new features or new marketing messages, the way companies do that right now is they do some polling, they might do some focus groups, they tend not to be very satisfied with the type of insights they get from those mechanisms. What we can do is basically build an artificial expert where we can, say, take a group of customers of a particular product and have them form this swarm intelligence and have that swarm intelligence give feedback about new marketing messages or new product features or even make predictions about how well those changes to products will do out there in the real world.
And so we’ve built swarms that will, for example, watch movie trailers and predict how well those movie trailers will drive people into the theaters. And these swarms of people can make more accurate predictions of movie box office than they could do on their own. And a really fun example of this, we did a few weeks ago, it was about a month ago, when the Oscars were happening. We were asked by a journalist to give them some forecasts for the Academy Awards and so we brought together a swarm of 50 people, just regular movie fans, and we had them, as a swarm, predict each of the awards in the Oscars. And then we were able to compare how well did this intelligence of regular people brought together as a swarm do, compared to industry experts. And what was remarkable is that the swarm intelligence was 94% accurate in predicting the Oscars, while the average professional movie critic was like 75% accurate. And this swarm of regular movie fans did better than the New York Times, than the L.A. Times, the Hollywood Reporter, Variety Magazine, because there is so much power in enabling this intelligence to emerge that’s greater than any of the individual members. And the individual participants who we had in this swarm intelligence for predicting the Oscars, as individuals, they actually didn’t do that well. And, in fact, we asked the participants, how many of you have actually seen all the movies? None of them had seen all the movies. In fact, most had seen, like, less than 1/3 of the movies that were up for Oscars. And so, as individuals, they were not that smart, but when they were working together as a swarm intelligence. As a system, they were 94% accurate in predicting the Oscars. And so it’s that type of intelligence that we’re currently providing as a business.
That said, as we look towards the future, we see that really any place where there would be value in picking a group of people and allowing them to make better decisions or better forecasts or even more creative solutions to problems, we see that swarm intelligence can apply. And we’re seeing interest from so many different industries. In fact, we have a project right now we’re doing with Stanford Medical School, looking at building swarms of doctors to see if a swarm of doctors can make more accurate diagnoses together than they could as individuals. And we suspect that they can because, again, nature has shown over millions of years that when organisms can work in systems, they can make significantly better decisions than they can on their own.
So how long have you had that system up and working?
So the first system that started building swarms of people was in 2014, so we’re going on 4 years.
And presumably, with the Oscars, you didn’t ask the swarm who should win. You asked the swarm who will win, correct?
That is correct. And that is always a really important distinction.
Right.
Is that you have to ask the question that you want the answer to.
So what would have happened if you put up the last election cycle? You put up all of the Republicans that were running for President, all of the Democrats, and you said, who should be President? And who will be President? Two completely separate questions.
Yeah, we actually—
Did you do that?
We did do that. We did that during the Republican primaries, and the Republican primaries were particularly interesting from a decision-making perspective because there were so many options, right? And the more options that you have, the worse traditional methods work in forecasting and also in reaching decisions. And so we very early in the Republican primary process, we started to have swarms of people predict who would be the Republican nominee for President and the predictions were coming out Donald Trump. And they kept coming out Donald Trump, even when the polls were saying that that likelihood was very low. And so this was now and year and a half ago and we actually started to wonder, Is there something wrong with our algorithms? Because we’re getting forecasts, they’re so different than the polls when we saw the group converging on the forecast of Donald Trump winning the nomination.
Now, once we got to the point where it was very close at the end, we actually asked the question that you just said, which is: What if we asked the same group of people not to forecast who will be the nominee, but to converge on who should be the nominee? And we actually got two different answers. So the swarms would say that Donald Trump will be the nominee, the swarms actually said that John Kasich should be the nominee. And that was a really interesting outcome. It kind of forecast the fact that the Republican Party was so divided. And because elections are done by polls rather than swarms, what an election will do is it will have the most popular choice emerge that’s not necessarily the choice that would maximize the satisfaction of a population.
And this is a great example of how the choice that emerged, while it was the most popular choice among Republicans, it wasn’t necessarily the choice that, had the group chosen John Kasich, it might have had the greatest overall satisfaction. And I can give an example of why this works this way with a much simpler question, and we do this with swarms all the time. A really simple question could be: Where should we go for dinner? Right? And I can give 5 or 6 choices where we should go for dinner. And I can say, “Let’s do it the way we elect people. Let’s just take a vote.” Right? And I can have one person vote for Italian food, and one person vote for Mexican food and one person vote for Chinese food and two people vote for Indian food. Well, that’s okay, two people voted for Indian food and there’s only one vote for everything else and so let’s go to Indian food—that’s the most popular choice.
Now, we have this presumption that that will maximize the satisfaction of the group, but there’s no reason to believe that that will maximize the satisfaction of the group because the group can go to Indian food and the other three people might hate Indian food. You might go and interview those people when they’re eating Indian food and find out that 3 of the 5 people are very unhappy because they hate Indian food or maybe they had it for lunch. Now, if it was a swarm, you’d have a very different outcome. In a swarm, what would happen is one person starts pulling the swarm towards Italian food and one person starts pulling the swarm towards Chinese food and one person starts pulling the swarm towards Mexican food and two people are pulling the swarm towards Indian food. And so the swarm starts moving to Indian food. But because it’s a swarm with feedback loops, everybody reacts. And so somebody who was pulling towards Chinese food will say, like, “Yeah, I wanted Chinese food, but I just had Indian food for lunch,” and they’ll switch the way they’re pulling. And somebody else pulling for Italian food might say, “I just don’t like Indian food,” and they might switch. And if this was a hundred people instead of five and they’re all switching, what happens is the swarm will start moving to Indian food for a second and then it will change directions, and it will start moving and it will basically find the path to the solution that the group can best agree upon. It will find the path to the solution that maximizes the collective conviction of this group. And so the swarm, instead of going to Indian food, it might go to Italian food. And Italian food might not have been anybody’s first choice, but if you interview that group when they ended up at Italian food, they will have significantly higher collective satisfaction than they would have had, had they just taken the simple vote and gone to Indian food.
And so this idea of we/us humans make decisions through polling, and it’s in a lot of ways far less evolved than the way that birds and bees and fish make decisions, which is through swarming. And a swarm will actually find that decision that really is the thing that reflects the combined sentiment of the population, whereas a poll is such an over-simplification, it’s really just what happens to be the largest plurality, but very often that’s very far from the answer that actually is the combined sentiment of a population.
All right. Well, let’s leave it there. The company is Unanimous A.I. It can be found at www.unanimous.ai. Louis, it was a fascinating hour. I wish you all well and it sounds like that’s a fantastic piece of technology you’ve developed. So thanks for being on the show.
Yeah, thanks for having me. It was fun.
Byron explores issues around artificial intelligence and conscious computers in his new book The Fourth Age: Smart Robots, Conscious Computers, and the Future of Humanity.