Collective Super-Intelligence For Decision-Making | Dr. Louis Rosenberg
📑 Chapters
00:00 Dr. Louis Rosenberg and ThinkScape
02:46 Meeting Others Where They Are
05:52 Swarm Intelligence in Nature
10:41 AI Agent Enhance Group Decision
17:17 Mitigate Dominating Bad Ideas
21:45 AI Agent Facilitate Large Conversations
26:10 Group Thinking and Collective Leadership
29:48 The Limitations of Polls and Surveys
32:41 Deliberation & Empathy in Group Decision
34:22 Joy of Building and Playing with Computers
36:50 Tackling Hard Problems and Building a Team
39:50 Technology-Enhanced Human Future
-
MAGICademy Podcast (00:00)
We're social spaces. We want to be connected in these large groups. we've been promised social media as this tool that connects people. social media really makes a lot of people feel isolated. It's not a real connection when you really do bring groups together and they can deliberate. They're having conversations. They feel empathy for the other participants in their group.
They really do feel connected and they really do converge on decisions that are good for the whole group because they feel part of the group. Biological systems have been evolving solutions to this problem for hundreds of millions of years. Whether it's schools of fish or flocks of birds or swarms of bees, they all have evolved methods of making really smart, really fast decisions together in groups. It's...
It's collective superintelligence. We are becoming smarter together, enabling large groups to be, to reach superhuman intelligence while keeping human values and human morals and human sensibilities and human interests inherently in the system. And that's really the danger from my perspective of artificial superintelligence. Artificial superintelligence will happen, but the risk is we don't know if a superintelligent AI
share our interests, or we'll share our values, or we'll share our morals, and that's dangerous. But if we can build a superintelligence that's based on lots of people, then we know it will inherently be human.
Jiani (01:47)
Welcome to MAGICademy podcast. Today we have a special guest, Dr. Luis Rosenberg, and he is a computer scientist, engineer, entrepreneur, and author. And his whole focus is in virtual reality, AI, and he started a company called Think-
scape, leverages AI to power group thinking in a non -biased and purely in powerful way. And he also founded many, many successful companies such as immersion corporations, for nursing simulations.
micro scribe, digitize, 3D
create, films like Shrek and Ice Age.
Louis Rosenberg, PhD (02:36)
Ice Age.
Jiani (02:38)
the first question would be, beep beep, a spaceship just land in front of you and out walks an alien. Very
kind, friendly, and very cute. And if you were to use one word or one sound or one movement to introduce yourself to this little friend, what would that be?
Louis Rosenberg, PhD (02:59)
Yeah, I would probably kneel down to get down to the aliens level. And I only say that because, as I mentioned, I've got lots and lots of animals here. I live actually on an animal sanctuary with chickens and ducks and goats and sheep and turkeys. And the thing that I've learned in interacting with every animal, especially when you're
when they're smaller than you is that if you get down at their level then they will interact with you as a peer as opposed to you know looking up and being being afraid.
Jiani (03:38)
I love that it's meeting them where they are as a foundation for the conversation to start. And I think this philosophy kind of eases into the think -scape that you're building, was harnessing the power of group work, meeting the groups where they are, rather than kind of coming from every other direction. Beautiful. Why Thinkscape
Louis Rosenberg, PhD (04:00)
Yeah, so, you know, my my whole career has been focused on the boundaries between technology and people and how to use technology to enhance human abilities. And and that's what got me interested way back at the beginning of my career on in virtual reality and mixed reality and augmented reality, because those are tools that can expand human abilities and and allow people to interact with information in
more natural and intuitive ways. over the years, I've become more more interested in not just using technology to enhance single individuals, but looking at how technology can enhance groups, groups of people. in particular, my interest has become for the last 10 years, how can technology enable groups, large groups, to make
better decisions and have smarter insights and basically become smarter together. like a lot of technologists, when looking at this problem, I look to nature as, how does nature do this? And it turns out that biological systems have been evolving solutions to this problem for hundreds of millions of years.
And so from an evolutionary perspective, you have a lot of different species, whether it's schools of fish or flocks of birds or swarms of bees. They all have evolved methods of making really smart, really fast decisions together in groups. And biologists call it swarm intelligence.
And it's swarm intelligence, whether it's a swarm of bees or a flock of birds. But let's just think for a minute about a school of fish, because school of fish is easy to kind of visualize. If you think of a school of fish, it could be thousands of members and nobody's in charge. And yet they can make decisions together in real time very, very quickly to navigate the ocean and make all of life's decisions.
quickly and effectively. so a lot of times, like when I say that to people, they say, well, school of fish, don't like, what kind of decisions do they have to make? imagine a school of fish and there's some predators coming at that school of fish. And there might be one predator coming from the left and another predator coming from the right and another predator coming from below.
Jiani (06:31)
Yeah.
Louis Rosenberg, PhD (06:45)
Only a few fish are going to see those predators coming. And yet in almost no time at all, that school of fish will make the right decision and move away from the predators and survive. so the big thing that interested me was, well, how do these fish do this? A school of fish is like an organization. It's like a big company of thousands of people.
And they're making these really smart decisions, even in situations where most of the individuals have limited information. And none of the individuals can see all three predators. Most of the fish in the middle of the school can't see any of the predators. And so how do they do that? And the question was, well, can we enable human groups to make decisions together that efficiently? And fish end up having this really amazing
Jiani (07:24)
Yeah, how do I do that?
Louis Rosenberg, PhD (07:35)
different types of magic. So one of the magical abilities of fish is that they have a special organ on the side of their bodies. It's called a lateral line, and it allows them to detect the vibrations in the water around them. And so each fish can basically communicate with the fish right around them. And by detecting the vibrations in the water, they know the direction and the speed that the fish around them want to go. And so
And so they have like a little conversation basically, like almost like a tug of war where they're pushing and pulling on each other to make a decision. And so fish, you know, on the left side of the school see a predator and they have a little tug of war amongst themselves and they decide, we should all, we should move to the right away from that predator. And then the fish all the way on the right side of the school see a different predator and they have a little conversation and they say, we should move to the left. Now that wouldn't solve the problem.
except that the other piece of magic that fish have is that these little groups of fish that can communicate, they overlap with other groups of fish that communicate because you have these little groups of neighbors overlap with other groups of neighbors overlap with other groups of neighbors. And so the information propagates through the school of fish. They all become very quickly become aware of the intentions of all the other fish, and they make this rapid decision. And and in almost no time, they will evade both predators.
even though the whole organization had different sets of information and have different intentions of what they think should happen. And so the question that drove my research was to say, OK, well, if a group of thousands of fish can make decisions that effectively, maybe humans could do something like that. And so humans are more sophisticated.
we don't communicate with by detecting vibrations in the water around us. We communicate with language. And so the question was, well, what if I took a thousand people like a school of fish? And I said, well, maybe I could break that break up a thousand people into little groups of, let's say, five people that can hold a conversation. And so I could take a thousand people and turn them into two hundred little groups of five people. And then nature would say, well,
If you can make all those little groups overlap, then information could propagate through the whole school. Right. And you could have a thousand people hold a single conversation. Well, it turns out that people are not good at being in two conversations at once. If I allow these conversations to overlap, you get you easily get confused. And in fact, there's there's a researchers have a name for this problem. They call it the cocktail party problem.
Because if you're at a party and you're talking to a small group of people at a party and there's another small group standing next to you and you start paying attention to that other small group, you lose the ability to pay attention to the group you're in. And so we humans cannot be in two conversations at once, which is why we've never been able to have a conversation of a thousand people at once by the way schools of fish do. And so we built this platform called Thinkscape that solves this problem.
Jiani (10:36)
Mm.
Louis Rosenberg, PhD (10:41)
And the way it solves the problem is it uses artificial agents, the type of agents like a chat GPT agent or it. And so it'll take a large group of a thousand people, break them up into little groups of five, and then it'll put an artificial agent in each of the groups. And that agent is there just to observe the conversation in the group and then pass insights to the agents in in other groups.
And then when the agent in another group gets an insight, it will express it conversationally. so this artificial agent weaves all the groups together into one conversation. so from a user's perspective, they're in a group of five people and one AI. And that one AI is not making up new information. It's not giving its own opinion. It's just relaying.
Jiani (11:21)
Hmm.
Louis Rosenberg, PhD (11:36)
insights that are being observed in other groups. And so information is propagating through all of the groups. And what we find is that we can have 500 people, let's say, have a single conversation. They could be anywhere in the world. We can give them a question from each person's perspective. They're just having a very calm, thoughtful conversation with four five other people and one AI. But the AI is linking all the conversations together and they can and the groups will reach a single decision.
The groups can brainstorm together. The groups can prioritize together. The groups can forecast or estimate. so what we find is that we can allow these large -scale group conversations, and groups get smarter, and they reach better decisions to hard problems. They generate more ideas. And very often, we see them reach decisions that are
more in the best interest of the whole group. They get decisions that are more holistic for the group as opposed to decisions that are very focused on just a few individual opinions. we see it as a really powerful tool for for enabling large groups to solve hard problems, for enabling large groups
to express their views as a population and find common ground, especially on issues where there's a lot of conflict. it's all based on how natural systems do it, how Mother Nature allows large groups to work together.
Jiani (13:16)
I love that. think whenever we get stuck, seek wisdom from the nature is always the most wise approach that we can ever take. also follow up questions. So according to research, five is the ideal number for the composition of like a number of participants in one group. Is that why is that? And research say like
Louis Rosenberg, PhD (13:26)
Sure, absolutely.
Yeah, so it's
a great, so, you know, one of the things that we, you we looked at from the start is, well, you can, one of the most powerful ways for people to solve problems is to have a conversation. And there's also a lot of research in collective intelligence that says large groups are smarter than small groups. And so you could say, well, if people work really well together in conversations and big groups are smarter, let's have a big, let's have a conversation of 50 people.
And it turns out that if you put 50 people in a room and you tell them to have a conversation, that's not really, that's not possible. So what's the ideal size? Well, researchers have looked at this and they found that five to seven people is a really good size to have a conversation. You have a good amount of air time per person and a small amount of wait time to respond to other people. But once you get to eight, nine, 10 people now,
each person gets to say less and everyone has to wait longer. And by the time you get to like 12 people, it stops really being a conversation. It starts to just be a series of speeches. And by like 15 people, it's a presentation, not a conversation. And so the research says the best size for a deliberation is five or six people. And that's why when we look at, OK, I've got 500 people,
If I can break the groups down to five or six people, but then use AI to connect all the groups together, then I get the benefits of both. The groups can have really thoughtful deliberations in small groups, but information and ideas and insights from all of the groups are spreading through the "school of fish." so you get the collective intelligence of a large group with the deliberation, the
the thoughtful deliberation of a small focus conversation.
Jiani (15:33)
love that. It's a beautiful way to scale without compromise quality. It's actually enhancing the quality because scalability brings. is that like 60 minutes minimum? And is that just one time? Does that happen maybe once every
Louis Rosenberg, PhD (15:53)
it depends on the group of people and the issue that they're going to discuss. Usually when groups come together and things get bits for 20 minutes to 45 minutes for just for conversation, most would be an hour and mostly because we want people to stay focused and more than an hour becomes people start to fade out. And then in terms of
You know how often they get together it could be you know there are groups that where they're going to come together once a month and other groups could come together every day they're going to have a conversation and things keep right now it's it's text based.
for now, text is the way that people like using it the best. And one of the reasons for text is that when it's text -based, people can be anonymous. And so now you can have 500 people.
having a conversation, everybody can be anonymous, which is actually really helpful and even in a large organization because it gets rid of the politics of the organization. People are not worried about, know, what does their boss think? Like they don't know who they're talking to. And so that gets rid of some of those dynamics. The other really interesting thing about conversations is that in any group conversation, you know, five people or 10 people,
a single strong personality can sway the whole conversation, right? There could be somebody who's just a loudmouth who pushes the whole conversation a certain way. What's really interesting is when you take a large group, like 500 people, and you break them up into these small groups, yeah, there could be a couple loudmouths, but they're only going to impact their small group of five people, right? They're not going to impact all 500 people. And if the loudmouth is pushing a
a bad idea, for example, in their small group, it's the AI will pass that idea on to other groups. And if those other groups think it's a bad idea, that it won't go anywhere. because it's structured as this swarm.
rather than a single large group. Also, you end up just with more ideas emerging because you have all these conversations happening at once. I very often in a large group, the first idea somebody says, if we had 100 people in a room, the first idea somebody says will influence all 100 people. Whereas if the group is now broken up, that first idea will influence a few people, but you have ideas emerging. And so you get
Jiani (18:26)
next time.
Louis Rosenberg, PhD (18:29)
you really get this power of, again, it's this biological power of swarm intelligence. The other thing about a swarm is that it doesn't always land on the most popular idea. Like if you take a, like it will land on what's the best idea because that idea has to propagate through the system and convince people. Because very often you can ask people a question and
they kind of have a gut response and everyone might have the same gut response. And so it seems like a really popular idea, but it doesn't really have good arguments behind it. But in a swarm structure, the ideas have to propagate and you end up converging on the smartest idea, not just the most popular idea. we actually did a really interesting experiment.
to test that. we did a project with Carnegie Mellon where we said, let's give groups of people an IQ test and say, let's measure their intelligence with an IQ test. And so we had groups, and these weren't huge groups, these were groups of 35 people. then we had other groups of 35 people who would just take the IQ tests on their own as individuals.
But then we would take the most popular answer from their responses. what we found was that if we brought these groups of people together, the average IQ of the individuals was 100, which is the way IQ tests are designed. If you had the people answer individually on their own tests and take the most popular answer, they boosted their IQ to 112.
which was like the 80th percentile. when the groups took the test together where they were deliberating and the AIs were passing messages around, it boosted their IQ to 128, which is actually the 97th percentile. So they were at the 97th percentile. And in fact, they were better than every single person who took the test.
It was able to turn this group of people into, you know, to perform at an intelligence level that was not just better than the average person, it was better than every person in the group. And so, yeah, so we've done lots of different studies that show that this structure, this, you think -scape using swarm intelligence, it makes groups smarter. It also makes groups solve problems faster and, again, generates lots more ideas. And so it's fascinating to me.
or we've done groups in the hundreds of people, but in theory we could do a million people. The technology is not quite there to do a million people at once, but you could. You could give a voice to a million people who would think together and solve hard problems together in real time.
Jiani (21:12)
Yeah.
Yeah, that would be like the ideal future Yeah, and then that's beautiful. And we explored the role of facilitators. Like with the traditional facilitator, there's still some sort of like power dynamics as like the facilitator kind of manages or owns the space and facilitate and time everyone and all that.
With the AI facilitator, you were mentioning that it's actually a new way of structuring small group conversations, a new way of facilitating. Can you explain a little bit more of how this AI facilitator differ from traditional facilitator?
Louis Rosenberg, PhD (22:01)
Yeah, so each of the groups has an AI that's participating. And when we first started working on the system, we did have the AI act like a human facilitator. And the problem was that people would assume it had authority, that they would believe everything that the AI said because it is in this role of authority. And really, we want the AI
not to have any special authority. We really just want it to participate in the conversation like anybody else. It just happens to be passing information that it's getting from other groups. And so now the AI that's participating in the group, it just acts like any other participant. It's just part of the conversation. It's introducing ideas. It's responding to things. But again, it's only based on what it hears from other people. We do have a facilitator that
to coordinate the global conversation. that right now is done by a human. So a human asks the question. A human would say, we just did a session with a large group of people who were asked questions about nuclear power. And so one of the questions was, are your biggest fears about nuclear power? It was 200 people.
And so the human, you ask that question, and then the human is watching as all the groups are having this conversation, and the AI agents are weaving all the conversations together. But as the group reaches an answer, the human facilitator would potentially say, have an idea for a follow -up question, and could then ask a follow -up question to the group. the structure really allows the, you
the person, whether it's a group manager or it's a market researcher or whatever it is, whatever that person's role is, to kind of sit back and allow the whole group to converge on ideas and decisions and then use those insights that get generated really quickly, like almost instantly. Like the group will have this conversation and after five minutes you'll see exactly what happened. And then the facilitator can then ask a follow -up question, either asking for elaboration or...
or asking, know, in that case of nuclear power, they first asked, what's, know, what are your biggest fears about nuclear power? Then they asked, well, what are the biggest benefits of nuclear power? And again, to see, what does the general public think about this issue? And it reveals really quickly these hundreds of different perspectives, and it finds which of the perspectives have the most commonality among
among large groups.
Jiani (24:53)
Beautiful, beautiful. There's like human and AI working side by side and being enhanced for each other.
Louis Rosenberg, PhD (24:59)
Yeah,
and the key thing is that the one thing that we really are careful not to do is that the the AI is not It's not giving its own opinions. It's not you know making up. It's not making up answers It's not going out and searching the internet for answers the AI the role of the AI is to connect all the groups together and so it's Observing other groups and then expressing those so it's it's weaving together the the human insights and the human ideas and it's allowing that
this large human group to efficiently have this large conversation and reach decisions, but it's not introducing its own perspective because first, the AI could be wrong. There's lots of times where the AI makes mistakes or introduces bias or other problems.
Jiani (25:44)
Yeah. Yeah, yeah.
That's interesting. That can almost deserve a different, like an independent topic about the risk of bias of the AI engine, how we can potentially reduce that. Curious, like moving into the future, what would be the best kind of scenario in terms of like group thinking or new form of like collective leadership?
with all the advancement of artificial intelligence, reality, I can definitely see Thinkscape being used in the virtual workspaces down the road sometime in the future.
Louis Rosenberg, PhD (26:22)
place that we say that the future is headed from this perspective is enabling large human groups to form a collective superintelligence. And so there's lots of AI research that pushes, it's pushing towards artificial superintelligence and
Jiani (26:33)
super.
Louis Rosenberg, PhD (26:40)
And that will happen, I believe, that it will become a time when AI systems are smarter than people in a large range of tasks and faster than people. that's really what most AI researchers are pushing towards is enabling these digital systems to have superhuman intelligence. The reason that we work on swarm intelligence is
to find this other pathway to superintelligence where it's collective superintelligence. We are becoming smarter together, enabling large groups to reach superhuman intelligence while keeping human values and human morals and human sensibilities and human interests inherently in the system. And that's really the danger from my perspective of
of artificial superintelligence. Artificial superintelligence will happen, but the risk is we don't know if a superintelligent AI will share our interests or will share our values or will share our morals. And that's dangerous. But if we can build a superintelligence that's based on lots of people,
then we know it will inherently be human. It will just be smarter than the individuals, but still share our interests and share our values. that's really the goal that we're headed towards. As we go, even long before we get to a true massive collective superintelligence, just bringing relatively small groups together, 30 people, 50 people, 100 people, 200 people,
we see that organizations can make better decisions, political groups can make better decisions, groups of scientists or engineers can solve problems better. And the thing about a swarm intelligence is that it really does find the solutions that a group can best agree upon. And that's really different than how humans right now solve big problems.
There's so many big problems where we can't agree with each other. we, the human groups can't decide what to do about climate change, can't decide what to do about poverty. There's different political perspectives and the groups become unable to make decisions. In nature, you never see that happen. A school of fish will not be unable to make a decision and break into two different schools of fish that go in different directions.
or entrench on their positions. Like that's just not how a swarm intelligence works. A swarm intelligence really does allow every participant in the swarm to understand what is the collective wisdom of that group, not what are these different factions in the group that are headed off in different directions and entrenching.
And one of the things that we've discovered in just looking at how do we humans deal with large groups right now, if you're a politician or a government, you want to understand the population, you very often take a poll. And the thing about polls are they're polarizing. What a poll does is a poll will show you the differences of opinion in a group. won't show you where the groups agree.
and it won't allow the groups to find common ground. But nature's method of swarm intelligence really does does that. What a swarm intelligence does is it allows the group to find the solution that best combines all of the insights. And that's why a school of fish can can have all these different predators coming and find it quite instantly, just find the right decision that's best for the whole group. And and so we think a collective superintelligence of people
will want to be smarter than individual people will be wiser than individual people because it be able to find the solution that's really best for the whole group as opposed to breaking up into factions that can't agree.
Jiani (30:56)
Yeah, and I think in traditional census, like lot of key stakeholders, for example, the policy is going to influence this and those people, and they don't get included. we're helping, same with the organizations, it's like key stakeholders are not being involved. having this ability to include as many people as possible and leveraging scalable small groups really can kind of at least give an opportunity for those voices and ideas to get into
the registration, get into the information, and then we can talk about it. I think I cannot speak for everyone from my limited perspective. I would assume, I would think that a lot of times when we reach disagreement, it's because we're missing some sort of important, like you said, common ground or perspectives. And there's no way.
Louis Rosenberg, PhD (31:46)
Right. That's the other thing that's
really interesting is that, you know, in large organizations, sometimes they'll send out a survey to all their employees to understand their perspectives. And then the management of the organization will make a decision based on the survey. And if the decision is not aligned with what you said, you will feel like, you know, I filled out the survey, but they didn't care what I thought. but when when this if you took the same group and you had them in a
Jiani (32:09)
Yeah.
Louis Rosenberg, PhD (32:14)
in a real -time swarm. They'll deliberate. They'll start to understand the other perspectives that people have. They'll have empathy for other perspectives. And so if at the end of the day, the decision is different than that person thought it should be at the beginning, it still might not be what they wanted, but they'll understand why it isn't. And they'll have more support. They'll have more buy -in for the decision.
Because it was an interactive process. It wasn't just filling out a survey and then finding out what the answer is later.
Jiani (32:48)
You feel you're part of the process to reach a collective decision. So it's very empowering.
Louis Rosenberg, PhD (32:53)
And what's
interesting is that when we first started working on this, we weren't sure if people would like participating in this swarm, a swarm like this. And it turns out people do like it. And I say that because, you know, we will bring groups together in a swarm and they'll be debating and deliberating and they'll be talking and then they'll get to the end of the session or they'll have answered the questions.
and people will stay and they say, ask another question, people want to be connected this way. Whereas nobody has ever been given a survey. And at the end of the survey said, I wish there were more questions on this survey. And so I think there really is something just biological about people is that people really, we're social species. We want to be connected in these large groups.
Jiani (33:32)
It's really the opposite.
Louis Rosenberg, PhD (33:47)
And we've been promised social media as this tool that connects people, but social media really makes a lot of people feel isolated, right? It's not a real connection. It's people passing notes to each other and clicking on upvotes and things. But when you really do bring groups together and they can deliberate, they're having conversations, they feel empathy for the other participants in their group.
Jiani (33:56)
Yes.
Louis Rosenberg, PhD (34:14)
they really do feel connected and they really do converge on decisions that are good for the whole group because they feel part of the group.
Jiani (34:22)
Luis, what did you enjoy creating or playing so much when you were around 11 years old that time just disappeared for you?
Louis Rosenberg, PhD (34:29)
Yeah, so when I was 11 years old is really when I first started working with computers and just writing video games. And this was back when that was rare to do when you were at that age. This was the very early days of computers. And I was just fortunate enough to stumble onto it.
And for me, it was just another form of play, right? Like you can, you could build with blocks or you could build with Lego or, and, or you could build by writing computer code to create, create things that I thought were fun. And, and that is kind of what I ended up always doing. Like for me, you know, building, building these software systems, I do it cause it's fun and it's interesting. you know, we, you know,
Jiani (34:59)
Yes.
Louis Rosenberg, PhD (35:25)
we just, at least with, with Thinkscape, we just keep being amazed with how human groups can get smarter and smarter and it just surprises us. And so it's, think for everyone on our team, we're doing it cause it's fun and interesting and it feels like we're, it feels like we're, it feels like play as much as work.
Jiani (35:49)
I love that. That's the ultimate... That's the best possibilities of work. Just like learning. It's not supposed to be sad or stressful or... Yeah, it's supposed to be play. I think we were born to play. What role does Childlike Wonder play in your life?
Louis Rosenberg, PhD (35:59)
Right.
Yeah.
yeah, no, think it's, you know, I, you know, I. People always say I work really hard because I like I spend a lot of time working, but I would say it's I it's because it doesn't feel like work to me. So that's so it feels it feels more like play, like if I'm doing it because it's I'm curious about something or I'm experimenting with something and it's to me, it's it is more like childlike wonder or play.
And so it's not, if it doesn't feel like work, then it's hard for me to say, well, yeah, I work really hard. mean, yeah.
Jiani (36:46)
You play really hard.
What do you think is your magic?
Louis Rosenberg, PhD (36:52)
is my magic. I mean, I do think that I work hard and from the perspective of again, doesn't feel like work, I will focus on an issue and really push it for a long time and work hard on it.
Jiani (36:53)
Hahaha
Louis Rosenberg, PhD (37:20)
And especially on issues, you I tend to get interested in issues that are where it's not clear if it's a solvable problem, right? It's, you know, and so it can take a long time, but we build a team of people that really are willing to willing to to go from ground zero to something that we know that there's this kind of amazing thing out there and work on that hard problem.
which is often hard to do in a startup environment or in company environment because there's usually pressure to get to a solution really fast. And that's maybe the smarter way to do a startup, but then to me that's a less interesting problem. It's a less interesting problem if it's completely clear that it's even possible when you start.
I tend to get attracted to the problems where this sounds crazy, but I think we can do it. And that's what we're doing now. You're trying to push towards collective super intelligence.
Jiani (38:20)
Yes.
I love that. That's beautiful. It takes a lot of courage and sharp insight to really see into the future and into the possibility and finding resources to potentially explore solutions. So, it's beautiful. It is magical and patient.
Louis Rosenberg, PhD (38:43)
Right. And it takes patience.
And you know that it's a long road. But I think, yeah, mean, you have to pick a problem where you know, OK, it's a long road. what you're trying to achieve is you're trying to solve a really big problem.
Jiani (38:52)
Somebody's gonna do it.
Yes, that's beautiful. And thank you, Louis for such an insightful, in -depth and happy and joyful and playful conversation and giving us hope and evidence -based hope for a future where we can actually thrive and develop and...
transcend our current limitations into a much beautiful collective future where everybody feels empowered and our human species as a whole can become more advanced in a technology enhanced way. Because that's the ultimate goal for technology, it's never to replace us, it's going to enhance us, make our human life much better.
make our planet earth happy.
Louis Rosenberg, PhD (39:51)
Yeah, yeah,
no, I agree. It's using technology to enhance what it means to be human, not to replace what it means to be human, which is one of my big worries is so because there's a lot of it would be very easy for AI to just replace what it means to be human. So.
Jiani (40:07)
Yeah, it may not
be sustainable, it can be destructive. I want to live in that universe. Thank you. I see Glenn is calling you. So thank you. Thank you, Glenn, for helping us to wrap up. He was not making sound until we kind of wrapping it up. He's like saying hi. So great. Thank you, Louis.
Louis Rosenberg, PhD (40:16)
Yep.
Yeah, he's back.
Yeah. All
right. Yeah, thanks.
💕 Story Overview
The MAGICademy podcast with Dr. Louis Rosenberg dives into the inspiring idea of boosting group intelligence by tapping into the wisdom of nature! It's all about creating spaces where people can truly connect and make decisions together, just like a school of fish moving in perfect harmony. Forget feeling lost in the crowd—this approach shrinks down conversations into cozy, manageable groups, while using AI to weave together everyone's brilliant ideas.
The goal? To spark empathy, create a judgment-free zone where everyone feels safe to share, and build communities that solve problems together, making sure every voice is heard and valued! It's about unlocking the amazing potential of collective wisdom, enhanced by technology.
MAGICal Insights:
Nature inspires enhancing collective decision-making: Examining how biological systems like schools of fish make rapid, intelligent decisions in groups can inform the development of technologies that improve human collaboration and problem-solving.
Small group conversations can lead to empathetic connections: When people engage in facilitated group deliberations, they often feel a genuine connection with others, fostering empathy and enabling them to converge on decisions that benefit the whole group.
AI-powered platforms can democratize conversations: AI agents can connect smaller groups within a larger collective, enabling large-scale conversations where diverse perspectives are shared anonymously, mitigating the influence of dominant personalities or organizational hierarchies.
Have you ever wondered how a large of of birds can fly into a therapeutic rhythm in harmony? Or how a school of fish will swim together in sync across a massive space inside the ocean while avoiding predators? No leader coordinates this response. No vote is taken. Yet somehow, this collective makes lightning-fast and life-saving decisions that benefit the entire group toward a goal.
This natural phenomenon—swarm intelligence—offers a profound blueprint for solving one of humanity's most pressing challenges: how large groups can make intelligent decisions from complex challenges together, almost simultaneously.
What is Swarm Intelligence?
"Biological systems have been evolving solutions to this problem for hundreds of millions of years," explains Dr. Louis Rosenberg, "Whether it's schools of fish or flocks of birds or swarms of bees, they all have evolved methods of making really smart, really fast decisions together in groups."
According to the research literature review (Liu & Passino, 2000), at its core, swarm intelligence represents the emergent collective intelligence of groups composed of simple “autonomous agents.”
“Unlike hierarchical systems where commands flow from leaders to followers, each autonomous agent in a swarm operates independently while interacting with its environment, which primarily consists of other nearby autonomous agents.”
This decentralized structure, perfected through evolutionary processes, enables sophisticated collective behavior without centralized control. Reynolds (1987) developed the "boid" model to identify a few basic rules that agents follow:
Avoidance: agents move away from those too close to prevent collisions;
Copy: agents align their direction with the general movement of neighbors;
Center: agents move toward the perceived center of the group to maintain cohesion.
View: where agents reposition when their perception is blocked (Flake, 2000).
Credit: Reddit
These simple principles—essentially forms of attraction and repulsion between neighbors—create remarkably complex and adaptive collective behaviors. A bird in a flock doesn't receive instructions from a leader bird (since no permanent leader exists), but instead continuously adjusts its position relative to nearby birds. This distributed decision-making allows birds to gain protection from predators (especially those in the middle of the flock) and to effectively search for food by leveraging the collective perception of the entire group, producing intelligent collective responses to complex environmental challenges.
Large Group Communication Challenge
In the podcast, Dr. Louis mentioned the role the fish’s “lateral line” plays as they navigate through turbulent circumstances, such as avoiding predators from multiple directions. According to Kasumyan (2003), among many functions, the lateral line system serves as a sophisticated biological sensor that allows each fish to precisely detect “movement” in water and perceive the “movement” intentions of nearby fish. As a result, the school of fish seems to move in cohesion fast without centralized coordination.
Fishes have this biological foundation for their collective intelligence. How about humans?
We face fundamentally different challenges. We communicate mostly by hearing languages and seeing visual signals. Even though we seem to function well in cocktail party situations by skillfully separating our voice from “noises” or direct attention to chosen topics, our ability to deeply process those conversations can be compromised as the group keeps getting bigger.
According to research, a natural way to process different information effectively in large groups is called Schisming (Egbert, 1997). Schisming describes a process by which a single group conversation naturally splits into multiple smaller conversations, often during social gatherings like parties. This happens when, for example, a group of people talking together gradually divides as subgroups form, each focusing on different topics or interests.
Schisming is a common social phenomenon and is part of how people manage the complexity of group interactions, allowing individuals to participate more actively and meaningfully in discussions that are relevant to them.
How can AI support this “schisming” process toward effective discussions in large groups, even on a global level?
A New Model for Collective Intelligence
Dr. Louis is exploring a way to collaborate with AI chatbots to power large discussions. The approach first divides large groups into smaller units of a maximum of six members (Bass & Norton, 1951) and harnesses AI chatbots intentionally to connect these small groups into one interconnected discourse system.
These AI chatbots observe conversations in each small group and relay insights between them, allowing information to propagate through the entire system, similar to how it moves through a school of fish. One important note from Dr. Louis is that:
“The key thing is that the one thing that we really are careful not to do is that the AI is not giving its own opinions. It’s not making up. It’s not making up answers. It’s not going out and searching the internet for answers.”
The AI's function is to bridge communication between all the groups by monitoring conversations in each group and sharing those observations with others. It essentially weaves together human perspectives and insights, enabling the larger collective to engage in an efficient, cohesive dialogue and arrive at more aligned decisions.
For example, in his study, Rosenberg, Schumann & Mani (2024) conducted with Carnegie Mellon University, a group of people took the IQ test on their own, and they got about 46% of the answers right. A group of 35 people took IQ tests together using this swarm intelligence structure and got about 81% of the answers right.
"It was able to turn this group of people into, to perform at an intelligence level that was not just better than the average person, it was better than every person in the group," Dr. Louis explained.
When we try to harness the large group intelligence to solve complex challenges, intelligence is not the only criterion. The best solution can appear not as the most intelligent but more thoughtful and inclusive.
Toward Collective Superintelligence
Unlike surveys, where no one asks for more questions, participants in these swarm-like deliberations often want to continue the collaborative experience.
"People really, we're social spaces. We want to be connected in these large groups. We've been promised social media as a tool that connects people, but social media really makes a lot of people feel isolated... But when you really do bring groups together and they can deliberate, they're having conversations, they feel empathy for the other participants in their group."
As artificial intelligence advances toward potential superhuman capabilities, this nature-inspired approach offers an alternative path: a collective superintelligence that enhances collective human thinking while preserving human values.
“The risk is we don’t know if a superintelligent AI will share our interests, or will share our values, or will share our morals; but if we can build a superintelligence that’s based on lots of people, then we know it will inherently be human. ”
As we look toward an uncertain technological future, nature's time-tested solutions for collective decision-making remind us that true intelligence may not reside in any single mind—human or artificial—but in the connections between them. This vision moves beyond current technologies toward systems that could potentially connect millions of people in meaningful deliberation, addressing our greatest challenges—from climate change to poverty.
Reference
Bass, B. M., & Norton, F. T. M. (1951). Group size and leaderless discussions. Journal of Applied Psychology, 35(6), 397.
Egbert, M. M. (1997). Schisming: The collaborative transformation from a single conversation to multiple conversations. Research on Language and Social Interaction, 30(1), 1-51.
Flake, G. W. (2000). The computational beauty of nature: Computer explorations of fractals, chaos, complex systems, and adaptation. MIT Press.
Kasumyan, A. O. (2003). The lateral line in fish: structure, function, and role in behavior. Journal of Ichthyology, 43(2), S175.
Liu, Y., & Passino, K. M. (2000). Swarm intelligence: Literature overview. Department of electrical engineering, the Ohio State University.
McDermott, J. H. (2009). The cocktail party problem. Current Biology, 19(22), R1024-R1027.
Reynolds, C. W. (1987, August). Flocks, herds, and schools: A distributed behavioral model. In Proceedings of the 14th annual conference on Computer graphics and interactive techniques (pp. 25-34).
Rosenberg, L., Willcox, G., Schumann, H., & Mani, G. (2024). Towards Collective Superintelligence: Amplifying Group IQ using Conversational Swarms. arXiv preprint arXiv:2401.15109.
Note: In the podcast, Dr. Louis mentioned the Cocktail Party Problem to refer to the challenge to process multiple conversations at the same time in large groups. However, the original research (McDermott, 2009) was focused on how humans are able to “tune in” to different conversations selectively in a gathering scenario and how difficult it is for machines to replicate the same process. We decided to use an alternative research on “Schisming“ to deliver what Dr. Louis was trying to convey.
⭐Louis’ MAGIC
Louis Rosenberg's "magic" lies in his approach to solving complex problems using the power of collective intelligence. Inspired by swarm behavior in nature, he aims to create digital spaces where large groups can tackle challenges more effectively than individuals. He is aiming to help people tackle complex problems that might otherwise prove insurmountable, ultimately leading to more holistic, well-rounded solutions that capture the wisdom of the entire group.
Connect with Guest
Louis Rosenberg, PhD, is a VR/AR pioneer and AI researcher with 30+ years of experience, starting at NASA and the U.S. Air Force, where he created the first mixed reality system. He founded Immersion Corporation (an early VR company) and Outland Research an (AR firm acquired by Google). Currently, he leads Thinkscape, focused on collective superintelligence. A Stanford PhD and former professor, Rosenberg holds 300+ patents and co-published "Our Next Reality."
https://www.amazon.com/Our-Next-Reality-AI-powered-Metaverse/dp/1399812246
Credits & Revisions:
Guest: Dr. Louis Rosenberg
Story Writer/Editor: Dr. Jiani Wu
AI Partner: Perplexity, Claude
Initial Publication: April 17, 2025
Disclaimer:
AI technologies are harnessed to create initial content derived from genuine conversations. Human re-creation & review are used to ensure accuracy, relevance & quality.
