Is AI Dangerous? Could Swarm Intelligence Be a Better Solution?

And it’s, already, creating biases, and other problems, but it’s gonna get worse. It’s gonna get worse because these AI systems are gonna get smarter and smarter. We’re gonna get more and more dependent on them. And they are not going to be the same as real human values, real human insights, real human sensibilities. And so my interest in this entire space was to say, hey, is there a better way to build a superintelligence?

Dr. Louis Rosenberg

Unanimous AI, a company owned by research scientist and a past guest of Association Chat, Dr. Louis Rosenberg, recently predicted the outcome of the 2019 Word Series using swarm intelligence.

The company didn’t guess right (few experts did!), but most of the time, they do. This is interesting for several reasons.

…if birds, and bees, and fish, can get so much smarter, together, why can’t people do it?

Dr. Louis Rosenberg

Some Background

Earlier this year a Twitter discussion featuring a debate over the “wisdom of crowds” caught my attention.

A friend of mine (and thought-leader in adult learning) was sharing why an unconference designed by the participants would always be better than a conference designed for the participants by one expert, or a small group of experts, who helped craft a vision for what a conference should be.

And it got me thinking about an interesting option that would turn the solution from an “either/or” to an “and” dynamic. And that option came from a segment I watched in an episode of a docuseries called This Giant Beast That is the Global Economy.

I remembered one segment, in particular, featuring a researcher talking with the host, Kal Penn, about collective intelligence, using a swarm intelligence technology, that could harness all the best wisdom from a group of people and make extremely accurate decisions.

Anyone who sees value in small data, big data, qualitative and quantitative research, any leader who makes big decisions for a group of people needs to hear about our topic today.

So I reached out to the researcher, Dr. Rosenberg, and his company, Unanimous AI.

This is a discussion about AI and humans that has a happy ending.

Or, if not exactly an “ending,” a happy pathway to offer folks who bemoan our future robot overlords and the direction AI-development seems to currently be taking us.

From our interview

Here is an excerpt from my interview with Dr. Rosenberg on the subject of AI and its potential dangers for humanity.

[Dr. Rosenberg] So, on the broader topic of AI, I am of the belief that AI can be dangerous. Which is surprising, because, you know, I am the CEO of an AI company.

[L’Italien] What are you doing?

[Dr. Rosenberg] …we are, really, a very different type of AI company. Most AI companies are looking at big data and they’re doing analysis of big data, machine learning on big data. So that they can, basically, replace people with algorithms.

[Dr. Rosenberg] They find patterns in these huge data sets, and they can replace people with algorithms. And that’s a little bit scary, because they are starting from this assumption that people are simple enough, in terms of their views, and opinions, and values, and insights, that they can be reduced down to these data sets. Then they process these data sets. Come up with patterns, that they can, then, use to replace people. And it’s, already, creating biases, and other problems, but it’s gonna get worse. It’s gonna get worse because these AI systems are gonna get smarter and smarter. We’re gonna get more and more dependent on them. And they are not going to be the same as real human values, real human insights, real human sensibilities.

[Dr. Rosenberg] And so my interest in this entire space was to say, hey, is there a better way to build a superintelligence? The traditional AI way is to say let’s replace people with algorithms, but if you look to nature, and you could say, hey, nature has addressed this problem, before. Where there’s been large groups of organisms, and they’ve found ways to, basically, combine their insights, and allow them to perform, together, as a superintelligence. And, again, it goes back to the birds and the bees. Flocks of birds, and swarms of bees, can be thousands of times smarter, together, than they can be, on their own, in terms of their decision-making capabilities.

[Dr. Rosenberg] And so, my focus, and the focus of the team, here at Unanimous AI, is to say, hey, can we use AI to connect people together, enable super intelligence, but not by replacing people, by amplifying people. And what we find is when we connect groups of people, together, they can be significantly smarter, they can make more accurate predictions, more accurate forecasts, more accurate diagnoses, have insights that are more generalizable, make better decisions, but, at the same time, they’re maintaining their natural values, and morals, and insights, and sensibilities, and opinions. We’re not replacing people. We’re allowing people to become smarter, together. And so, to me, that’s a much safer approach, for pursuing superintelligence, by keeping people in the loop, as opposed to what a lot of researchers are doing, which is finding ways of replacing people with straight algorithms.

Instead of just looking to algorithms, let’s connect people together. Amplify their intelligence. And we can get the power of AI, while still maintaining the real value of human wisdom.

Dr. Louis Rosenberg

What if you could harness all the best wisdom – the knowledge, experience, and intuition – from a group of people and use it to make extremely accurate decisions and predictions? Even more accurate than artificial intelligence (AI) by itself? Would that be valuable for you? If the answer is yes (and how could it not be), you must watch this interview with the CEO of Unanimous AI, Dr. Louis Rosenberg.

What could you do with swarm intelligence? How would it change the way you make decisions? These are the questions that will endure beyond the World Series.

---

Full Transcript

Message from this episode’s sponsor:
Support for this episode of Association Chat comes from Boldr Strategic Consulting. You can drive your business strategy with a digital plan, that unites and engages your team. Boldr Strategic Consulting can bridge the gap, between your technology and marketing departments. You need insights to inspire business transformation. Boldr will empower you to thrive in a digital world. You don’t have to stress out about your digital transformation projects, anymore. Let Boldr Strategic Consulting come in, and help you, and your team, reach success. Find out more at boldrstrategy.com.

• [KIKI] What if I told you there was a new way to make important decisions, that could impact every industry, everywhere, and could even change the way we look at how we conduct elections? I was really looking forward to this interview with Dr. Louis Rosenberg, and for me, at least, it did not disappoint. Immediately following the live interview, which, as I’m recording this, was just yesterday, I received emails from people thanking me. I even received one email, from someone who wants me to be at their annual meeting, when they put the technology, we talk about in this interview, to use. I am not kidding, incredible stuff. And this was one of those interviews that I really am thankful that I had the chance to do. And I’m also looking forward to more work, coming from the wonderful researcher that we interview in this episode, Dr. Louis Rosenberg. Alright, so, here we go folks. A few weeks ago, I saw a tweet from my friend, Adrian Segar, who, literally, wrote the book on unconferences. And he was discussing the wisdom of crowds sharing why an unconference would always be the better choice to a conference designed using a conference curator, or one expert who helped craft a vision for what the conference should be. And this created, a sort of, a controversial discussion online, and it got me thinking about an interesting option, a sort of yes, and, and, moment. And that came from a segment that I watched in an episode of a new Amazon Prime Video docuseries. It was just released, February 22nd, called This Giant Beast That Is The Global Economy. And I remembered one segment, in particular, featuring a researcher, talking with the host, Kal Penn, about collective intelligence, using Swarm AI technology, that could harness all of the best wisdom, from a group of people, and make extremely accurate decisions. And, I wondered, could this technology be used to make better conferences, or for associations of every industry, facing tremendous change and disruption, could Swarm AI, or hive intelligence, help make strategic decisions about an organization’s mission. Anyone who sees value in small data, big data, qualitative, quantitative research, any leader, who makes decisions for a group of people, needs to hear about our topic, today. So, I reached out to the researcher, and his company, and our guest, today, received his bachelor’s, master’s, and PhD degrees from Stanford. And his doctoral work focused on robotics, virtual reality, and human-computer interaction. As a researcher, at the US Air Force’s Armstrong Labs, he created the virtual fixtures system. The first augmented reality platform ever built. Our guest has been awarded more than three hundred patents, for his technological efforts. Dr. Louis Rosenberg is CEO and Chief Scientist of Unanimous AI, a company focused on amplifying the intelligence of networked human beings, human groups, using AI algorithms modeled after natural swarms. And, today, I’m happy to welcome Louis to the show, to talk with us, more, about these questions, and so, so, many others. Welcome to the show!
• [LOUIS] Hi, thanks for having me!
• [KIKI] I’m so excited! I really, really, have been looking forward to this, and was kinda proud of myself for, you know, remembering, and looking, and finding you, from that show. But, I went through that intro really quickly, and I wanna make sure we start with a definition of what we’re talking about, today. So that people truly understand what this is all about. So, can you walk us through swarm intelligence?
• [LOUIS] Sure, so, in the natural world, there is really two different types of intelligence that has evolved over millions, and hundreds of millions, of years. There’s neurological intelligence, which is basically what we think of, in our brains. It’s networks of neurons that are so deeply connected, that with billions and billions of neurons, in your head, intelligence emerges from this dense network of connections. Most people think of that as being the only form of intelligence, in nature, but biologists have discovered that there is another level of intelligence, which is swarm intelligence. And that is, it’s evolved, usually in social species, species like birds, and bees, and fish, where they have these large groups, large organizations, you could think of them as associations. And they have a large group of individuals, they’re all out there, in the world, getting different perspectives on their environment, collecting different information. And what swarm intelligence does is say, hey, can a group combine their various perspectives, in an optimal way, and actually connect together, so deeply, that they function as a superintelligence. And biologists refer to this as a superorganism. And so, out there, in the natural world, a swarm of bees can make remarkable decisions, by functioning as a superintelligence. A school of fish can make remarkable decisions, by functioning as a superintelligence. Same with flocks of birds, and, in fact, countless social species. And it comes down to this fact, that they can leverage all their unique perspectives, and become significantly smarter together, than they could be, on their own. And my interest came from this fact of, hey, if birds, and bees, and fish, can get so much smarter, together, why can’t people do it?
• Right.
• We didn’t evolve the capability, naturally, but, what we’ve been working on, here at Unanimous AI, for the last five years, is to say, hey, can we use AI technology as a networking technology, as to connect groups of people, together, over the internet, and allow them to have the same benefits. And it turns out that we can. And we can make groups of people significantly smarter, by having them connect, together, into these systems, modeled after natural swarms. And I should say that, from a biologist’s perspective, the word swarm refers to all of these types of organizations. Whether it’s a school of fish, or a flock of birds, or a swarm of bees, a biologist would refer to that as a swarm, and the emergent intelligence, as a swarm intelligence.
• I mean, it’s fascinating, because associations are known for associating. We’re known for bringing people together. And we have our own swarms, you know? And so, figuring out how to empower humans, so that we’re able to better use these communities that we already have, just seems like a natural progression. Seems like a natural thing that we need to figure out how to do. And I wanna go into that a little bit more, but, you know, you spoke about we haven’t developed this naturally, as humans. And we’re social animals. We need community to survive. We haven’t developed it naturally, but we’re developing it through Unanimous AI, and the work that you’re doing. And so, what that makes me think of, is all of these human versus machine, human versus AI, these dire predictions about how, you know, we’re sunk. Because we’re creating something that’s gonna take over, and the AI’s going to be smarter than we are. What I found fascinating, in some of the other interviews with you, and your TED Talk, is about this idea that maybe not! Maybe, in this case, humans can come out ahead! Can you talk to me, a little bit, about that?
• Sure. So, on the broader topic of AI, I am of the belief that AI can be dangerous. Which is surprising, because, you know, I am the CEO of an AI company.
• What are you doing?
• It’s really because we are, really, a very different type of AI company. So, most AI companies are looking at big data. And they’re doing analysis of big data, machine learning on big data. So that they can, basically, replace people with algorithms. They find patterns in these huge data sets, and they can replace people with algorithms. And that’s a little bit scary, because they are starting from this assumption that people are simple enough, in terms of their views, and opinions, and values, and insights, that they can be reduced down to these data sets. Then they process these data sets. Come up with patterns, that they can, then, use to replace people. And it’s, already, creating biases, and other problems, but it’s gonna get worse. It’s gonna get worse because these AI systems are gonna get smarter and smarter. We’re gonna get more and more dependent on them. And they are not going to be the same as real human values, real human insights, real human sensibilities. And so my interest in this entire space was to say, hey, is there a better way to build a superintelligence? The traditional AI way is to say let’s replace people with algorithms, but if you look to nature, and you could say, hey, nature has addressed this problem, before. Where there’s been large groups of organisms, and they’ve found ways to, basically, combine their insights, and allow them to perform, together, as a superintelligence. And, again, it goes back to the birds and the bees. Flocks of birds, and swarms of bees, can be thousands of times smarter, together, than they can be, on their own, in terms of their decision-making capabilities. And so, my focus, and the focus of the team, here at Unanimous AI, is to say, hey, can we use AI to connect people together, enable super intelligence, but not by replacing people, by amplifying people. And what we find is when we connect groups of people, together, they can be significantly smarter, they can make more accurate predictions, more accurate forecasts, more accurate diagnoses, have insights that are more generalizable, make better decisions, but, at the same time, they’re maintaining their natural values, and morals, and insights, and sensibilities, and opinions. We’re not replacing people. We’re allowing people to become smarter, together. And so, to me, that’s a much safer approach, for pursuing superintelligence, by keeping people in the loop, as opposed to what a lot of researchers are doing, which is finding ways of replacing people with straight algorithms. Just to give a sense of how powerful this is, I can give you some examples. One recent example that I think is particularly interesting is we had a collaboration, we have an ongoing collaboration with Stanford Medical School. The interesting thing about the medical field is that there’s actually a lot of doctors who are afraid about AI replacing their jobs.
• Yeah, yeah.
• You would think, of all the jobs that are out there in the world, doctors would be the least afraid, but, in certain parts of medicine, doctors are on the front line of potentially being replaced by machines. And the doctors who are most afraid are radiologists. Radiologists, they look at X-rays, and CAT scans, and MRIs, and there’s been a bunch of researchers who have created straight algorithms that can process these images and do just as well as human doctors. And so, Stanford Medical School came to us and said, hey, let’s see if we can take the crown back, for humans. Instead of replacing doctors with algorithms, let’s see if we can connect, together, small groups of doctors, in swarms. And so, we have a software platform we call Swarm. And Stanford Medical School did a study where they had small groups of doctors, six doctors to eight doctors, who logged into SWARM, and then they would diagnose chest X-rays, together, as a system. And so, the chest X-rays would pop up on their screens, and they would diagnose whether or not the patients had pneumonia, or not, as a swarm. And what we found is that, when they were working together, in the Swarm software, and we can talk about how that works in a little bit, they’re, basically, connected together by AI. The study showed that they reduced their diagnostic errors by over 30%. So, they had 30% fewer errors, when diagnosing chest X-rays, when connected together as a swarm, as compared to how they would perform, individually. Also, they compared this against the straight algorithms, and this group was 20% more accurate than the straight, machine-learning, algorithms. So, it took the crown back for humans.
• Yes!
• That’s really, kind of, the approach that we take. Which is to say, hey, people are smart. People are really smart. People have knowledge, and wisdom, and insight, and intuition. We take it for granted, but it’s massively powerful. We have so much experience in our heads, and, instead of just looking to algorithms, let’s connect people together. Amplify their intelligence. And we can get the power of AI, while still maintaining the real value of human wisdom.
• I mean, that’s the part that I think is so fascinating, because there’s so much about this that is, but that, I mean, how do you look at wisdom? And figure out how to quantify, how do you make an algorithm out of wisdom? You know, like, what do you do? And, you know, a little bit of this, and a little bit of that. You know, hesitation here, and confidence there. How do you do that? It’s just fascinating!
• Right, so, the interesting thing is that we’re not trying to build wisdom, we’re trying to capture wisdom. Meaning, the humans in this system are bringing their wisdom, and what we do is we allow groups of people to connect together, as a system. And so, if you think back, you know, this idea of getting the most out of a group. Let’s say I have a group of 20 people, and I wanna get the most out of their insights. And they all have different perspectives, different experiences. They might have different views on the world. If they’re business people, they might have different specialties. And so, they all have really valuable pieces of information. How can we combine them, in the best way? Well, the traditional way that has been done, sometimes people refer to it as the wisdom of crowds, which is to take a vote, or a poll. I can take a poll or a vote, people can give their input, and when you do that, you do see the group get a little bit smarter. This wisdom of crowds effect works, but it’s using techniques that are, you know, a hundred years old. And there are some big issues with just taking a vote or a poll, which is that you are assuming, when you take that vote or poll, that everybody’s feelings have equal levels of conviction. But, really, people have very different levels of conviction, and confidence. And so, what we do, in Swarm, is we, instead of taking a vote or a poll, we create a system. And the system, a lot of people say it looks a little bit like a Ouija board.
• You know, it makes you wonder about some of these old, you know, where it’s like, how did it know? And it’s like maybe we were working on this, early.
• But what happens is a question will appear on everybody’s screen, at the same time. And that’s the thing about a Swarm, is it’s synchronous meaning everybody’s participating, together. You know, a vote or a poll, people will just give their data, in isolation, as individuals. You know, you call it a crowd, but there really isn’t a crowd. It’s just a bunch of individuals, and then, the crowd exists in somebody’s spreadsheet, somewhere, when they find an average. A swarm is really a group that is coming together. They’re logging in at the same time. They’re all, questions appearing on all their screens, at the same time, and then, they have an interface, where they, basically, a question appears on all their screens. A bunch of answers appear. And they have a little magnet, where they can pull the swarm in a direction. And everybody’s pulling in different directions. You have this, kind of, multidirectional tug of war. People pushing and pulling. And then, the AI algorithms are watching all their behaviors, how they’re pulling against each other. Because they’re basically interacting, and the algorithms are saying well, who, you know, which people are very confident in their opinions, which people have very low conviction. And there might be a variety of different options. So, which people are confident, with respect to some options, but, maybe, ambivalent, with respect to others. And, with all that happening in real-time, the swarm starts moving in a direction, and it finds the path to the answer that maximizes the collective conviction, of the group. And the results are, always, really, really, powerful. So, I can give you another example. So, one of the things that we do a lot is we take on challenges from journalists, and Newsweek challenged us to predict the Oscars, this way. And, I should say, we as a company, we don’t know anything about the Oscars. We don’t have a big data set about the Oscars. We just know we can amplify the intelligence of people. So, we invited in, to the platform, 50 movie fans. Not experts, just regular movie fans. All they need is a browser. They can log in from anywhere in the country. They log in to the Swarm platform, and then we ask them to predict each of the different categories of the Oscars. And so, the question pops up on their screen, who’s gonna win Best Actor? The actors appear. And then, they basically swarm, and they converge on an answer. So, we did this for all the categories. We gave the results back to Newsweek, and then Newsweek published the predictions, which puts pressure on us. And we were 94% accurate, in predicting the Oscars, which outperformed every major movie critic. It outperformed the LA Times, the New York Times, Variety, Vanity Fair. It outperformed all of the experts. What happened is we took just 50 regular people, combined their insight, and they became a super expert. An expert that was better than the specialists at the LA Times. And here’s the most interesting thing is that we, then, asked those 50 people, well, how many of you have seen all the movies? None of them had seen all the movies. In fact, most of them had seen less than half the movies.
• Wow.
• But they’re out there in the world. They heard different things. You know, some people read articles in the newspapers. Some people heard something on the radio. Some people saw some of the movies. So, they all have different pieces of incomplete information, but when they come together, as a swarm, they’re filling in the gaps in each other’s knowledge. The system is seeing, well, these people were very confident about these particular actors, while these people were confident about those actors. And it allowed the system to, basically, form this amplified intelligence. This artificial expert that was better than the actual experts. And we see that time, and time, again. And we can do the same thing for businesses, or for associations. You can imagine if you are, if you have a large community, whether it’s a community of enthusiasts, or a business team, and they all have different views on the world. They’re out there, maybe, in different parts of the country, or different parts of the world, and you wanna be able to combine their insights, in an optimal way. That’s what a swarm can do. It can, basically, allow you to combine all their different perspectives, and find that answer, that is the best combination of their insights.
• I love it. And, actually, we have someone who’s commenting about how you could change the entire betting dynamics, on the Triple Crown, using this tool. And, yeah, I think you’ve been in contact with some people, about that, yeah.
• Yeah, so, we’ve done a lot of sports forecasting. And it’s not because we’re interested in sports, it’s because it’s a great testbed, to demonstrate amplification of intelligence.
• Right.
• It’s the type of thing where we can get a group of people, just regular sports fans, they can log in and they can make a set of predictions. And then, we can rigorously look at how did they do. And, in fact, we just published a paper, where we did the full NBA season, we predicted over 200 basketball games, during the season. We had people predicted, as individuals, and, then, predicted together, as a swarm. And then, we gave the different groups a simulated betting, basically, simulated dollars, to see how they would do. And what we found was that across these 200 games, the individuals would have lost 41% of their money, by placing the bets on their own, individual, intuition. But when they combined their intelligence, as a swarm, they averaged 170% gain, against Vegas.
• Wow.
• And so, we see this with allowing groups to have this, to be able to make better judgments, make better decisions, whether it’s diagnosing a chest X-ray, or betting on a basketball game. And that’s one of the things that’s really interesting about it, is that the same exact software works, whether it’s a group of doctors, or a group of sports fans, or a group of movie fans, because the algorithms are just looking at human behaviors. They’re not looking at, there’s no information, you know, we’re not processing information about medical content, or movie content. We’re looking at human behaviors, and optimizing how to combine their insight.
• Well, so, let me just say that imagine that you’re at the Thanksgiving table, and you’re looking around the table, and there are some people that, you’re thinking, might drag down the intelligence quotient of your swarm. You know what I’m saying? And so, I can imagine that people out there are wondering, you know, what goes into getting good results, with this? If you’re really trying, I’m thinking of associations, specifically, here, but anyone, who’s really trying to get good results, does it matter, on an individual basis, how you’re figuring out who’s part of this swarm? You know?
• Right, so, the group, it does matter who the group is. But it really depends on the questions you’re asking. So, I gave you the medical example. If I took a group of regular people, off the street, and asked them to diagnose chest X-rays, they’re just gonna be guessing. And so, there could be no amplification of intelligence. On the flip side, there’s a lot of questions where just regular people, off the street, are really insightful. And so, we do a lot of projects, and we have a lot of companies using the software for doing things like product insights, marketing insights. In fact, we did a project for a large fast food chain, where they wanted to predict how TV ads would do. And so, they would bring in groups of their customers, into the platform, and then, the actual TV ad would pop up on all their screens, and they would watch the ad. It was actually an advertisement for pancakes. And then the swarm would get questions about, you know, will this advertisement drive people to go to the restaurants? And you could ask a bunch of really interesting questions. And so, that’s a situation where what you’re really thinking about is just saying, hey, we can take a group of customers, turn them into a swarm. They become an artificial, expert, customer. And then, you can interview that expert customer. Or you can create an artificial expert medical doctor, by getting a group of doctors, and they could make more accurate diagnoses. But, in both cases, you’re picking the right population of people, to come into the swarm. Now, you had started out by saying, well, what if you have somebody who, maybe, you know, either doesn’t know very much, or they have… And that definitely happens. What we found is that it doesn’t matter if we have a group of experts, or a group of novices. What happens, in all cases, is that the swarm will be smarter than the individuals, on their own. And so, when we look at the difference, meaning I can take a group of regular people, and they can turn into a swarm, and they can give me more accurate insights, than they would have done on their own. Or I can take a group of experts, and they will become a super expert, and give more accurate insights than the experts would have done, on their own. When we look at individuals to say, well, who are the best participants? Is it the person who knows the most, that’s the best participant? Turns out, it’s not the person who knows the most. It turns out, it’s the person who’s the most self-aware.
• Oh, okay.
• It’s the people who know what they know, and know what they don’t know. Those are the best participants. People who are, really, overconfident, meaning, they think they know more than they do, they, actually, are less effective contributors, than people who might, maybe, know less, but know that they know less. And we did a really interesting experiment, And this was with the university. I’m trying to think of the name of the university. Marist, they have the Marist Poll, like, one of the big polling universities. And we did a study where we looked at sports fans, versus sports writers. Who could give more accurate forecasts? Fans versus writers? And, it turned out, the sports fans, actually, gave more accurate predictions, in a swarm, than the sports writers. And the reason was the sports writers were all, extremely, overconfident. They thought they knew more than they did.
• Wow.
• And because they’re sports writers, they’ve trained themselves to always look for the very unusual things to write about. So, they never predict the most likely thing to happen. They predict something that’s a little bit unusual, because nobody would read the articles, if they…
• Otherwise. Right, yeah, yeah yeah!
• They actually were not as effective. And so, that’s a situation where the sports writers probably knew more about sports, than the fans, but the fans knew what they knew. Knew what they didn’t know. They weren’t overconfident, and they became good participants.
• Fascinating! Okay, ’cause what does that mean then? There are so many things that come up, with that. What does that mean, then, for those of us who are thinking about our boards, or our panel of experts, for our associations? Who would be the best, in the room? Does that mean we, then, need to look at, maybe, HR, and say, have these people taken assessments that say how self-aware they are, you know? Maybe we don’t want the people who are overly confident, in the room. I don’t know! What does that mean?
• Yeah, I mean, it’s interesting, because what we’ve found is that people who are skilled at what they do, tend to be self-aware. And I say that because one of the things that was interesting, to us, is that we’ve done lots of different projects and swarms, with people who range from novices to real experts, like medical doctors. I mean, the radiologists at Stanford, you know, they went to school for 12 years to become radiologists. They’re at the top of their field. They’re at Stanford Medical School. You might think, well, maybe they’re all gonna be arrogant, or all gonna be overconfident, but, really, they’re all professionals.
• Yeah.
• They know what they know, know what they don’t know. We did a project with Boeing, where Boeing had swarms of military pilots. And this was a really interesting project. See, Boeing’s interest is to design cockpits for military aircraft. And they are always trying to get feedback from pilots. And they normally give surveys, or focus groups, or do interviews. And Boeing, in collaboration with the US Army, did a study where they had groups of military pilots form swarms. And Boeing published a paper about it, where they said the swarms gave them much more accurate insights about the feedback about cockpits, than they would get by doing a focus group. And the reason for that, is that these, again, these pilots are probably the best at what they do. They probably, if you met them, you might think they’re arrogant at what they do. But they’re professionals, and they know what they know, and they know what they don’t know. And they were actually very effective at swarming.
• Okay, well, I wanna give us a second to answer a couple of questions that have come in. One came in, and it says, “how are swarms identified “as viable predictors, say, “how would a market react to a product?” So, okay, so I think I understand what this person is asking. What do you have to say to that? How well are swarms used to predict how a market will react to a product?
• It’s used, a lot, in that way. It’s one of the more common uses. We have projects with some of the largest companies, in America, where they’re trying to predict how a new soft drink might be received by the markets. And they can bring in swarms of customers, and make predictions. Most of those projects are confidential, but there’s one that we just did where we were actually able to write it up as a case study. And it’s actually on our website. This was just done. There’s a media group called Bustle, Bustle Media Group. They’re the largest publication for millennial women, in America. And they wanted to do a test to see can they predict the sale of clothing, in this case, it was during the holiday season, just in December of 2018. And so, they did a study where they had small groups of millennial women come into the Swarm platform, and they looked at a set of eight different sweaters. And they were asked to predict which of these sweaters would be the top sellers, and which would be the bottom sellers. And what they found was that these swarms of, these small swarms, they were each about 15 people, were able to predict, with extremely high accuracy, which would be the top sellers, and which would be the bottom sellers. And they compared this to just a traditional survey, and the survey, basically, couldn’t predict, at all, which were the top sellers, and which were the bottom sellers. Again, the case study is actually on our website, but it was a really good example, because they allow us to, publicly, talk about it, and then they actually had the sales results, in January, they actually had the sales volumes that came out of the Christmas season. So we could see, yeah, these groups were able to very, very, accurately predict sales. We’ve seen other groups predicting supply chain, and inventory issues. We have done a project with a major TV network, to predict which new fall TV shows would be successful. And, there, they had swarms of people watch the trailers for the TV shows, and they were able to accurately predict which shows would get renewed, which would get canceled. Again, what you see is that people, people are smart.
• Yeah.
• They have really good, not just wisdom, but just intuition, about, not just their own opinions… And that’s an interesting thing, because we don’t ask people, you know, what’s your own opinion about this TV show, or what’s your own opinion about that sweater? We ask people, let’s predict how the public will react to this TV show.
• Interesting. So, yeah.
• Because we’re treating the individuals, not as data points, we’re treating individuals as data processors. We’re saying, look, I could take 20 people, but they each, they know their own feelings. They know their family’s feelings. They know their neighbors. And so, we can ask them to, basically, they’re like the bees, out there in the world, collecting information. And they’re coming back, and then we can amplify that information.
• Yeah, yeah! They’re not saying where I want my hive, they’re saying where would be a good, you know, where would everybody want to build this hive, right?
• And it’s one of the really big differences between, say, your traditional market research, where you take a poll, and you just ask people their individual opinions, and a swarm, where you can use much, much, fewer people, but you’re treating them as data processors. You’re treating them, saying, each of these people has been out there in the world, has collected huge amounts of information, knows lots and lots of other people, let’s have them work together, as this collective intelligence, and forecast, you know, will these product features be successful? Or will this marketing message resonate with the public? And what we find is that a group of 20 people, in a swarm, can outperform a survey of hundreds, or even thousands, of people.
• Wow, I mean, that’s really incredible, too. And, I think that, you know, I’ve seen a couple of questions come in about the actual technology involved, and making sure that you’re getting the best types of results, that you possibly can. One of the early ones, which I have to credit to Randy, here. And she’s a smarty, I’m proud of her. Like, immediately, she was like, okay, was talking about the machine biases into algorithms, and how do we deal with the challenge of human nuances? And you’ve talked a little bit about how you’re framing the questions, so that people are not just answering, as individuals. And you’re, also, looking for things like overconfidence, that feeds into this algorithm. What are some other things that help to make sure that the responses, or the answers that you get, are the high quality, and more accurate, responses, that you’re looking for?
• Yeah, so, it’s really a great topic, and something that we think a lot about. What we find is we can bring in a population, and we can get really good insights. One of the good things is… An ideal population has diversity of perspectives. Meaning, if I approve people who all think exactly the same, then there’s not much that’s gonna get amplified, right? So, we have a group of people who, already, have the exact same opinion of what the right answer is, then that’s the answer that’s gonna come out of a poll, or a survey, or a vote, or a swarm. If you have a group of people who, actually, have conflicting views, who, actually, have strong opinions, but they have diverse opinions, then, in a swarm, you’ll, actually, see this combination of insights that allows the best [Inaudible 00:39:59] to emerge. And so, one of the things that is most interesting, is, especially, when you have a large association, right? When you have people, maybe, people from all over the country, or all over the world. They can log in, form a swarm, and they, usually, become a very powerful swarm, because they have these diverse perspectives.
• Yes. So, I’m going to explain why, if you’re watching this, right now, we, suddenly, have a third person, on the screen. If you’re listening to this, later, then it won’t even matter, but if you’re participating in it, live, you’re probably wondering who is this lovely person who just joined them?
• Who is this interloper?
• And I wanna introduce Tracy Betts. She’s the CEO for Boldr Strategic Consulting. Tracy, you work with associations on digital transformation projects. Projects that have a lot of big questions connected with them. And projects that demand a lot of big decisions being made by leaders, in charge. So, I asked Tracy to come on, and, actually, ask a couple of questions, as a representative, for the show’s audience. Hearing all of this conversation, so far, what are a few questions that you have for Louis, that you’ve been thinking of?
• Yeah, so, Dr. Rosenberg, you answered the most important question, to me, which is, I am famous for being with full conviction, believing the BS that comes out of my mouth.
• Look, I am right!
• I have a question as far as, okay, how does that play into… But I do have a deeper question there, as how are you determining conviction levels?
• Right, so that’s another really, really interesting, so, one of the problems with traditional methods, like just taking a survey, is that people are very bad at reporting how they feel. You can ask people a question, you know, which product feature do you think is the best, and how strongly do you feel? And you could ask them on a scale of one to ten, how strongly do you feel about this? And, a, people don’t necessarily even know how they feel, but, b, even if you give me a number eight, on a scale of one to ten, and somebody else gives me a seven, we have no idea if your scales are the same. My internal scale and your internal scale are different. And so, as a swarm, we don’t ask people to report, we just get people to behave. And by behave, like I said, all the people are interacting. They have these little magnets, and, they’re basically, it’s like a multidirectional tug of war. And so, the algorithms are watching how they behave. Are they pulling for an answer, and, quickly, switching to something else, when they realize that the swarm is going in a different direction, or are they resisting, and being entrenched? Or, maybe, they switch from one answer, to a different answer, in defiance of the direction the swarm is going. And so, as the swarm is behaving, or as the swarm is moving, everybody is revealing a lot about themselves, just by how they’re interacting. And that’s really what the algorithms are looking at. It’s not asking people to tell us how strong their conviction is, it’s just asking people behave naturally, as a group, and those things are emerging, from their behaviors.
• Interesting, thank you, yeah. So, I read an article about how WPP was using your technology. It was the study that they did with, I think, they took two groups, of 40 participants, to try to think about what the impact of, the future of digital in 2022, on the retail industry.
• Yep.
• Right? Okay, and what I thought was really interesting is they had these two swarms, one representing the west, right? North America, Europe, South Africa. And the other representing the east. And there was a big divergence in what the two swarms came up with, which was fascinating to me. And if you apply that to associations, like, if I look at, you know, our associations are trying to deal with global issues. So, if we take the issue of ocean pollution, right? It’s a big issue for the society of Naval and Marine architects, or engineers. It’s a big society for the plastics industry association. I’m sure American Chemical Society is looking at it. But, if I took a swarm from a group of engineers, versus a swarm from a different group, I think I would come up with a very different answer, and, maybe, one that wouldn’t be conducive to truly solving a world problem. So, how, thinking about a global issue. I have two questions here. The first question being how is it that you, if that was posed to you, what would that group look like? First part. And then the second part is I’m very curious about scaling, as far as how big, is there a sweet spot for the size of the swarm?
• Right, so, well let’s answer the size question, second. But this issue of different populations is really interesting. If you did build a swarm of engineers, and asked them questions about a global issue. And you built a swarm of philosophers, and you asked them the same question, you would get different answers. Because, in [Inaudible 00:45:50], you’re building, when you build a swarm of engineers, you’re building an artificial expert engineer. And with philosophers, you’re building an artificial expert philosopher. They’re gonna be different. And that’s, actually, interesting. You might actually want to know what are the perspectives of these different groups. If you want to get a more general answer, then having a more diverse group is better. So, again, if you’re trying to get this general type of solution, getting a diverse group, often, surprises us, about how how you can get these very, very, selfless, moral decisions. We did a project with Fast Company, about a year ago. So, about a year ago, Jeff Besos had tweeted out requests for what should he do with his fortune? Like, for philanthropy. And he tweeted out, what should he do? And he got, like, 50 thousand tweets came back of all these different ideas of what he should do with his fortune. And so, we did this effort, with Fast Company, where we took the 50 thousand, and processed it, and there was a lot of overlap. And it came down to just a few hundred, different, ideas. And then we had a swarm of just people from the general public, who went through these couple hundred ideas, and narrowed it down, and picked, you know, what would be the best use of the funds. All the participants were Americans. And there were things on there that would be very attractive to Americans, like free healthcare for all Americans, was like a suggestion. Free medication for Americans. And, yet, what the swarm came up with was clean water, for everyone in the world.
• Wow.
• Which was really interesting, because it was one of the choices that really, maybe, had the least benefit to this population, who are Americans. Whereas, so many of the others, maybe, would have. And then, like a few months later, the UN had their list of top priorities and clean water was one of the very top. And so, what we saw there was that you could take a group of people and have these insights come out that it’s not just people picking in their self-interest, they really are giving answers that are best, for the group, as a whole. And that’s another thing we see about swarms, is that they get people thinking about things in a different way, because they really are part of a community. Whereas, if you give people a poll, or a survey, a poll or a survey is actually very isolating. You’re giving a few pieces of data into a black box, and then, if the answer comes back, and it’s not what you put into the black box, you think, well, nobody cared. Nobody cared what I said. But in a swarm, you’re actually part of the process. And so, even if the swarm goes to an answer that, maybe, you didn’t have as your first choice, you have more buy-in, because you are part of the process. To get to your second question, you asked about the size of the swarm. It’s a really good question, something that we look at, a lot. When we first started, we assumed you needed 40 people, 50 people, at least. We’ve looked at swarms of hundreds of people. We’ve seen academics do research studies, which show that very, very small groups, even groups as small as three to five people, as a swarm, can be statistically significantly smarter, more accurate. And so, it could be a business team, as small as three to five. Stanford Medical, those were radiologists, six or seven radiologists. Boeing, when it was military pilots, those were four or five military pilots. Or we’ve seen groups of 50, 100, 200. It really depends. Our recommendation is if you have a group of experts, then a small group is pretty effective. We would recommend, you know, five to twelve experts, and you can get really, really, accurate results. If you’re looking at a group of consumers, or a group of people who have a broad, then 50 to 70 is a really good number. We’ve found that if you go from 70 to 100, there’s really not much of a benefit. So, there’s diminishing returns, when you start going higher, and higher. The one place where we do see an interest in doing large groups, and we haven’t mentioned this, and I know you have an interest in conferences, live events.
• Mmhmm.
• One of the things that we can also do is we have the ability to have groups swarm, in live events, where they use their phone, instead of a computer. And so, the swarm is up on the screen at the front of the room. Everyone could pull out their phone. There, if you have 300 people in the room, you could have 300 people in the swarm. You don’t, necessarily, need to have all 300, but part of it is just getting engagement of the whole crowd.
• So, I just have to say that, yeah, that’s awesome, because you just answered a question that we had over here on YouTube. Ron, who is with a very innovative, healthcare organization, in Chicago, was just asking, can you set up swarm activities at annual meetings? Can it combine both virtual and in-person participants?
• You, actually, can. I, actually, think it’s most interesting when you do that. When you have a group of people, in the room, and then there’s also people, just out there in the world, who are also part of this swarm. And so, you see this swarm up on the screen, and it could be people… But, we absolutely can have live groups swarm. We’ve done some interesting projects. The XPRIZE Foundation did a live event, where they had a bunch of their, they called them, visioneers. They’re experts and they were trying to decide what future contests that they should fund. Again, going to one of the world’s big problems that they wanna solve. And they used swarms, live. We, on a, kind of, completely different end of the spectrum, Credit Suisse, last year, they had their Asian investor conference. Where they had people from all over the world, who are experts in Asian markets, and they actually ran swarms, up in the front of the room. And they were, actually, forecasting what are the important trends in Asian markets that they should predict. Right there, live, in the room. Actually, the Credit Suisse example is also a case study that we have, on our website.
• Oh, okay, that’s great! I have seen a couple of questions that have to do with, I think, probably the question that you must get, every single time, and that has to do with politics. And so, so, what does this mean for, could we use this for deciding things, like political decisions, or voting? And so, how do you answer questions like that?
• So, I believe that there’s a lot of value in using swarms for political decision making. And I’m a believer that polling is a big problem, for politics. Polling, my view is that polling is polarizing. What a poll does is it shows the differences in a population. It, actually, then reinforces those differences, and so the polling process gets groups to entrench. Swarming is really the opposite. Instead of revealing the differences in a population, it reveals where groups can agree. And it, actually, then, reinforces the places where groups can agree. In nature, for example, honeybees have these difficult [Inaudible 00:54:00]. They have to decide where they’re gonna move their hive to, and, if the group’s entrenched, the way we humans do, for politics, they wouldn’t decide, and that colony would just die out. And so, I’m sure, over millions of years, all the colonies that had this tendency to entrench, died out. And so, now, that never happens with honeybees. They, always, can reach a decision. We could learn a lot from that. The swarming process can get rid of gridlock, can get rid of this entrenchment, and allow groups to converge. We, actually, just started a project. We got a grant from an organization called Nesta, which is a big, British, philanthropic organization. And we’re actually looking at Brexit. And looking at groups of voters in the UK, and can they agree on what’s the right solution, for them to get out of their Brexit mess. It’s very hard for them to reach a decision that everyone can get behind. And there’s a lot of different options, so it’s, actually, a really interesting type of decision. But every country faces similar things.
• So, this is an important one. KiKi, can associations afford this? The swarm would be awesome as part of our FutureThink exercises. Where do we invest our innovation fund to make the biggest impact? Yeah, so, what are the costs associated with looking at doing a swarm? Having a swarm?
• So, the process is, actually, relatively affordable, partly because we just launched a new [Inaudible 00:55:48] Platform. People can use the software, themselves, and it’s priced, in some sense, very similar to tools like Slack, where it’s by user.
• Okay.
• Like ten dollars, per user, per month. And then groups can use it as much as they want. Whether they’re asking one question, or they’re asking thousands of questions. Our goal is to make it affordable, and just get groups using it. And there’s other ways of doing it, where it could be just a certain number of guests. Instead of having named users, it could be a certain number of guests. But the pricing is on our website, and it really is intended to make it accessible to all organizations, big and small.
• I’m sure everyone loves hearing that. Tracy, I want to take a second to say thanks, so much, for adding your questions, and providing some additional perspective. I’m really trying to incorporate this into the Association Chat, so that we can have a moment where we can collect questions from other people, and, kind of, have our own representative, of the swarm, come on, and this was a perfect chance to do that. So, thank you.
• Thanks.
• Alright, so, I guess we are getting close, and I wanted to ask you to maybe take a second and share a vision for where this could go. You know, the thing is I know that you’re constantly working on making progress, and developing, and seeing new applications for it. So, what would the future look like? And, as we learn more about swarm intelligence, and this human and AI powered decision making, where might that take us?
• Yeah, it’s a great question, because I feel like we’re just at the beginning of this. This is a technology area, where we’re seeing real amplification of intelligence. We’re seeing groups able to make forecasts and decisions that are 20%, 30%, 40%, more accurate. I suspect that, five years from now, we’ll be looking back at that and saying, you know, that was just the early steps of what we can do. And, again, it goes back to this idea that people are really smart. And the level of knowledge, and intuition, and experience, that people have, really, it’s untapped. And really focus on how can we leverage that, how can we combine it, and how can we think about a business team as this collaborative intelligence that can be amplified. And so, my hope is that as this technology matures, people will realize that it’s really important to keep the human part of decision making active. And not take this, kind of, this trend toward just replacing people, and automating decision making, with algorithms. Again, I’m a firm believer in the power of AI,
• Yeah.
• It’s just, you can’t lose the human part of it. Partly, because you can’t capture, really, what, you know, the opinions, and values, and sensibilities, effectively, in an algorithm. But, partly, also, that just people know a lot. We can just get better decisions. We can get better forecasts. We can get better insights, by having people be part of the process.
• I love this idea that, the idea that, somehow, we can capture that mysterious thing that it is, to be human. And some of that intelligence that’s there, that the closest thing we can do is call it, what, intuition? And yet, you know, how do you account for that in a way that is practical, and makes a difference. And when you combine it with the intelligence of other beings, you know, other humans, I think it’s just a really fascinating way to think about the way we make important decisions. So, any final words of wisdom? And, if people wanna find out more about your research, or maybe how to talk with you about speaking to their groups, or hosting a swarm, where can they find you?
• Sure, so, our company is Unanimous AI. People can find us just at unanimous.ai And, at that website, the software is called Swarm. People can go to the website. There’s actually a two week, free trial, so people can try it with their group, with their team. There’s online help, so people who have questions… And then, if anyone has questions about the stuff we’ve talked about today, if they send you questions, you could, certainly, email me, and I’m happy to email back. Our interest is getting groups to realize that they can take advantage of this technology. And associations are great.
• Yeah.
• You have large groups, and they’re diverse. And one of the things that can happen in a small business team, is they can get into an echo chamber. Where they all have the same views. An association is interesting because that breaks that. You have people who are from different organizations, different companies. They have different views. And then, how do you combine them? So it’s a really good use.
• I’m telling you, any association person who’s listening to this, right now, knows we’re in the middle of spring conference season. This would be a perfect time to start figuring out how you can pull something like this together. If not by when your conference is happening, or your annual meeting is happening now, then in the fall, when fall conference season starts. And so, I wanna say thank you, so much. I am so thankful that I not only got a chance to see what you did on that series, but then, you know, your team is fantastic. You’re fantastic. And just taking the time to talk with us, today, I think it’s incredibly important. And I wanna say thanks to Tracy, and everyone else who’s watching, right now. I hope you’ve learned something helpful, today, that’s going to help you, and help your organizations, wherever you are. And please join us, next time. Until next time, everyone! Keep asking questions to learn, every day. As Joseph Campbell once said, “the cave you fear to enter, “holds the treasure you seek.” You have a great week, everyone. Thanks for listening to this episode of Association Chat. If you like it, please subscribe. Or, better yet, tell a friend! You can join the Association Chat book club, or support Association Chat by visiting Association Chat’s Patreon site. You can also find more information about live events, private communities, special projects, and more, on associationchat.com. See ya next time. Hey, Association Chat is in its tenth year, and independently owned and produced by me, Kiki L’Italien, and a very small crew of freelancers, and volunteers. We appreciate our sponsors like you wouldn’t believe. So, I want to give a special thanks to all of our sponsors. Today, that includes Boldr, EventWaves, and Amplified Growth. Thanks to all of you. And if you want to find out more about sponsorship, then go to associationchat.com, or email me kiki@amplifiedgrowth.net, and we’ll be happy to talk with you! Alright, thanks!

Resources