Nicholas Christakis: [00:00:05] All right, so human beings are embedded in social networks, and these networks obey very particular mathematical, biological, and social principles. And increasingly, we are adding artificial intelligence in the form of online agents and physical robots among us who interact with us as if they were social entities. And these sorts of agents that we're going to be adding to our systems range from driverless cars on the roads to checkout machines in stores, to humanoid robots in homes or in factories or in battlefields or in firefighting situations, to disembodied autonomous agents such as online bots and digital AI assistants in our phones or our eyeglasses or our workplaces. And these technologies interact with us on a level playing field as if they were human. And they will give rise to hybrid systems of humans and machines. And these systems offer opportunities for a new kind of social artificial intelligence.
Now, let me give you just a toy example of this. When you get, like, a digital assistant, like an Alexa, the manufacturer of that device is very concerned with the human-machine interaction. And that human-machine interaction is optimized. For example, you would never buy the Alexa if every time you needed something from it, you had to say, excuse me, Alexa, I'm very sorry to interrupt you. If you don't mind, would you please tell me the word? [00:01:35] weather tomorrow, right? This would be considered an absurd level of politeness. You expect to be able to say, Alexa, weather, and then the machine obediently responds. And that's fine until you bring that machine into your house and your children talk to that machine and learn to be rude. And then they go to the playground and they are rude to other children. So that machine that has been added into our midst, it's not just about the human-machine interaction, it's about the human-human interaction in the presence of machines. And so what I'm interested in is not, is that, is the human-human interactions in the presence of machines. And we can use an understanding of social network structure and function to assess the uses and impact of social AI within and upon human groups. with respect to factors such as trust and cooperation that are necessary for groups of people to work together, uh, and that affect the behavior of these collectivities.
So let me highlight some of the work that we're doing, uh, using several approaches to network experiments involving, uh, artificial intelligence. These experiments evaluate how AI might affect the structure and function of human social interactions. [00:02:45] Now there's a class of collective action problems of social interactions that we have that are known as coordination problems. And that is, these are problems in which we have to work together if we were to create something of use. And sometimes we solve these problems by creating centralized institutions like police or courts or governments. But often we are able to solve collective action problems that require us to coordinate or cooperate with large numbers of people in a decentralized way. We actually evolved to have this capacity. And one specific type of collective action problem that I'd like to start with today is, in fact, coordination.
So, for example, to avoid this traffic jam, people have to coordinate to do something dissimilar from their neighbors. So if everyone leaves their house at the same time, everyone is stuck in a traffic jam. But if, if they jostle their departure times and they leave over intervals, then nobody is in the traffic jam. Now, of course, you could have some kind of central authority that coordinated this. You guys leave first, then you, and then you. [00:03:45] But ideally, what you would want is some kind of decentralized, non-top-down way of human beings coordinating to solve this problem.
So here was our first experiment that was published in 2017 on such hybrid systems that explored how AI might help with such a challenge. And we explored the performance of human groups that were engaged in a coordination task. And our paradigm was to borrow from computer science something known as the graph coloring problem. This is a classic problem in computer science. But what we did is we took it and we put human beings in that situation. So what we did is we took 4,000 people and we put them in 230 online groups. So these human beings were dropped into these groups. And they were randomly assigned to a spot within an artificial network that we created, the structure of which roughly resembled real human networks. And they were dropped into these spots and they were told they were assigned one of three colors, [00:04:46] a purple, orange, purple, orange, and pink. And they were told that they had to pick a color dissimilar from their neighbors.
They were allotted a brief period to engage in the task. Individuals observed their surroundings, noting the colors of their neighbors. Each person considered their situation and determined the necessary adjustment to achieve a distinct color. Conflicts arose when adjacent individuals shared the same color, indicated by a red line. Background lines depicted the dynamics of social interactions. Participants would check their neighbors periodically, making choices in order to differentiate their color from others, all while aiming to work collaboratively for compensation. In this experiment, time was represented along the x-axis up to the five-minute limit, as shown on the y-axis with the number of color conflicts illustrated. Initially, a number of conflicts emerged due to random color assignments, depicted in a histogram. As participants interacted and adjusted their colors, they encountered a situation where two individuals experienced a conflict deemed unresolvable. Resolvable conflicts were indicated by light orange lines, highlighting individuals who could adjust to alleviate their conflicts. Conversely, certain participants found themselves unable to shift their color to reduce conflicts within their immediate environment. This stagnation resulted in a collective impasse, requiring a surprising choice by one individual to temporarily increase conflicts in order to promote progress. Eventually, after 245 seconds, a solution was reached, prompting the system to end the game and dispense rewards.
Due to our clever approach, we conducted experiments where we subtly substituted some human participants with bots. We assessed how the integration of AI-enhanced bots into a hybrid system influenced group performance. Can we incorporate bots into human teams to enhance their coordination capabilities when confronted with such challenges? We introduced three bots and varied two factors experimentally. We positioned the bots randomly in the network, either placing them at the core or on the edges. Additionally, we manipulated their AI capabilities in a straightforward manner, deciding if they should operate flawlessly or with a degree of randomness.
In the flawless mode, every 1.5 seconds, the bots observed their neighbors and chose the color with the least conflicts. This could be perceived as illogical behavior. In the 10% randomness scenario, they followed the same process but occasionally selected a random color, specifically 10% of the time. In the 30% randomness scenario, they operated similarly, but with a random choice 30% of the time. Hence, we increased the likelihood of errors, making the bots increasingly unpredictable.
Next, we examined a control group. On the x-axis, we plotted time, while the y-axis represented survival curves. This reflected the probability of the group collectively failing to complete the coordination task. Initially, at time zero, all human-only groups were represented in orange, showing that 100% of these groups had not resolved the task. Over time, however, more human groups succeeded, leading to a scenario where, after five minutes, around 60% of the human-only groups had managed to complete the task.
What occurs is that when you introduce a small percentage of noisy agents in a central role within the network, you observe a notable enhancement in performance. In this scenario, a significantly larger number of groups of individuals, upon the introduction of these slightly noisy agents to the central hub, managed to resolve the issue. Specifically, they reduced the average time taken to find a solution from 232 seconds to 103 seconds. Additionally, there are other insights gleaned from this data. Both perfectly functioning and excessively noisy agents proved to be detrimental. A certain level of adjustment was necessary. It was the small group of noisy agents that proved to be the most advantageous, and the location of these agents also had an effect. Importantly, in these experiments, it was noted that individuals who were not linked to the agents, who were positioned further away within the network, underwent changes in their gameplay as well. The data illustrated some individuals connected to the agents and others who were not. It was observed that even those isolated individuals began modifying their approach. This created a ripple or cascade effect. The positive interactions between the agents and human players permeated the network, subsequently influencing interactions among humans further afield. In essence, the agents facilitated self-assistance among the humans, and the advantages derived from the noise spread throughout this social framework.
Now let me further fix ideas regarding this collective challenge and how these simple AI agents might help with another analogy. So imagine you have a plane. So this is like in gradient descent in machine learning, for example. So you have a plane, and you have hills and a mountain. And you have different hills, okay, of different heights. And you have a mountain way up here that's the tallest mountain. So you drop, I'm going to take four of you, and I'm going to drop you somewhere here, and I'm going to handcuff you together, each of you looking in a different cardinal direction, [00:10:54] and I'm going to blindfold you. And I'm going to say to you, find the highest mountain. So you guys talk among yourselves and you say, well, why don't we each take a step in our direction and report back to the team? So you take a step north, and you say it's uphill from here. And south says it's downhill from here. And east and west say it's lateral from here. So you all agree, let's take a step north. And you keep doing this iteratively until you reach a point when all of you say it's downhill from here.
Have you found the highest mountain? No. (...) What have you done? You found the nearest hill.
Now will you ever find the highest mountain? No. You will never find the highest mountain. You are stuck. You are locally optimized, but globally sub-optimized. And in order to globally optimize, you need a little noise. You need occasionally to allow this group of people to make a counterintuitive step down the mountain or step down the hill. So they take a sequence sometimes by chance of a sequence of steps until they get back onto the plane. And then they navigate all around this fitness landscape, exploring all these peaks until they wind up on this peak. [00:11:56] And this tall peak tends, the global optimum, tends to be a receiving state because it takes much more noise to get off that peak than the other peaks. And so now you oscillate around the global optimum. So in our work, we've been exploring this kind of simple programming inserted into social systems to see if we can improve the performance of human beings in addressing diverse kinds of collective action problems.
Now another collective action problem involves a different challenge, which is cooperation, not coordination. Humans often have to cooperate to produce what is known as a public good. And a lighthouse is one of the canonical examples of a public good. Public good has two canonical features. First of all, it's what's called non-excludable. And that means that other persons cannot be prevented from using it. If you build a lighthouse for your own personal sake, because you're navigating the seas and you don't want to crash into the shore, that's great for you, but you can't stop anyone else from using it, okay? Non-excludable. [00:12:58] And also, it's non-rivalrous. That means that consumption by one person does not reduce consumption by others. If I use the light from my lighthouse, there's no less light for you to use. And this is unlike, for example, a piece of cake. If I have a piece of cake, it's mine, right? I can prevent you from eating it. And if I eat it, there's none available for you, okay? So public good has these features, and it's these features that in turn make it very difficult to produce public goods. Because when it comes to building a lighthouse, it's very tempting to pit your individual interests against the groups. If you don't contribute to building the lighthouse, you can still benefit from it. And so everyone is tempted to do nothing, and then the lighthouse doesn't get built to the detriment of all. And it's also worth emphasizing that public goods are useful, because you can actually produce things with them, such as safe sea travel. And as such, the underinvestment in public goods is a serious problem in our society, and has also come to be known as the tragedy of the commons. These public goods, for example, norms of trust that we maintain among us are efficient. [00:14:01] Think of when you were at a high school. Some of you went to a high school where students trusted each other, and that meant you could leave your backpacks alone in the hallway and not worry that anyone would steal them. Others of you went to high schools when there weren't the same norms of trust, and now you had to lock up your backpack or keep it with you at all times. In which of those two environments do you think you would have better learning? Right? In the former environment. So that norm that's collectively maintained is productive. In that case, productive of learning. Here, productive of safe sea travel and so on.
So collaboration among individuals in groups for the creation of shared benefits is complex, necessitating various strategies to maintain it. We have conducted numerous experiments where individuals are placed into networked groups to engage in different types of public goods games with their peers, altering multiple structural factors over time. For instance, we initiated an experiment years ago in which individuals were introduced to their neighbors and participated in a public goods game drawn from behavioral economics. In such a scenario, I might contribute a small amount of money to each of my neighbors, like taking a dollar and distributing it among them. The scientists then double the contribution. For example, if I have three neighbors and I give a dollar to the collective, it becomes two dollars, which is shared among us four. Thus, the entire group benefits by two dollars, while I personally receive only 50 cents, the result of two dollars divided by four. This means I have to make a personal sacrifice for the benefit of others. Naturally, many individuals tend to resist this, thinking, 'Why should I sacrifice? Let someone else contribute.' However, if everyone adopts this mindset, the system collapses. The most advantageous scenario occurs when everyone maximally contributes. In this experiment, we commence the game, and the blue dots symbolize the cooperative individuals who generously contribute to their neighbors, facilitating public goods like constructing a lighthouse. On the other hand, the red dots represent the exploiters, also referred to as defectors, who don’t contribute. Interestingly, our findings in this experiment echo a result established over three decades ago: cooperation tends to break down within groups. By the end of the game, after multiple rounds, virtually everyone turns into a defector, save for a few blue individuals who sustain civilization among themselves. Think back to high school, where your science teacher assigned you to a group project with three other classmates, leading to shared grades. You might find yourself partnered with less motivated peers, leading to two options: either you do all the work while they also receive good grades, or you decide against contributing and all receive failing marks. Typically, individuals opt for the second choice, attempting to avoid feeling taken advantage of, which results in failure to cooperate within social systems more broadly. To address this issue, we introduced different types of bots into the system—simple entities serving as social brokers. These bots observed local interactions and provided advice, suggesting individuals sever ties with defectors and connect with more cooperative peers instead. This gentle reorganization, based solely on local knowledge without any authoritative enforcement, demonstrated in our trials with over 1,000 participants across 64 groups, that not only could cooperation be preserved, but with these bots present, cooperation levels could actually rise beyond initial measures. DeepMind later replicated and expanded upon our findings in another paper published about a year or two hence.
In a different experiment, we examined the impact of bots on group creativity. [00:18:07] finding new ideas can be challenging. Theories and experiments indicate that groups might excel in recognizing and sustaining innovations when they share their discoveries.
However, group innovation encounters specific obstacles, one of which is groupthink. Correct? When a group of people is gathered, they can quickly settle on an inadequate idea.
Imagine, if you will, that a group might possess greater collective wisdom and could generate more ideas. For instance, if tasked with enhancing a fishing rod, the first individual might suggest adding a hook to a string. So, one person now holds the string with the hook. Then someone proposes attaching a stick to the string, which is seen as a brilliant idea. They merge their insights and proceed. One might observe that the bait on the hook is hovering on the water's surface and suggest incorporating a wave to submerge it. However, that causes it to sink too deep, leading to the proposal of a bobber for tracking its position, and so forth. This collaborative innovation leads to collective wisdom and the preservation of knowledge over time, [00:19:08] resulting in cultural artifacts that represent a blended output of multiple individuals sharing concepts and being creative.
So we wanted to create a game in which groups of people searched a landscape for an optimal idea. And we decided to use nouns as a proxy for ideas. And we took 20,000 nouns from the classic computer science word-to-vec corpus. So we took 20,000 nouns. And the distance between these nouns could be defined by the cosine similarity metric. You can imagine a hyperdimensional vector space in which cat is more similar to dog than it is to desk. And the way they did that is they looked at how often the words cat and dog co-appeared on websites. So they had a universe of websites and a universe of 20,000 nouns. And they said these two nouns oftenly co-appear and these two other nouns don't often co-appear. They created a 300-dimensional vector space. [00:20:08] And now you can describe how similar are any two nouns. And we decided to use nouns as a proxy for ideas. We took these 20,000 nouns. And then we picked a set of nouns. Imagine we picked one noun, but we picked a set. One noun at random from all of these, like braggadocio, for example, was a noun. So we picked braggadocio. And we say that's the perfect idea that we want this group of people to find. And then all the nouns that are near braggadocio fall away in this vector space. So you have the peak noun that gets the most points, 20,000 points, and all the other nouns to the nouns farthest away. And we put human beings into this system and we say, find this word. We don't tell them the word. And then we tell them the point value of the words. So they start guessing. And as they guess, they get feedback. And they say, ah, this word has more points than another word. And they get closer and closer and they start sharing the knowledge with each other. They're trying to be creative to solve the problem.
So each noun is related to others in a semantic space or a fitless landscape. [00:21:10] And we had 18 different target nouns, as if we pulled a peak up from the landscape of these 20,000 nouns in 18 different locations in 18 different landscapes. And these nouns were deliberately unusual, like fratricide, shoehorn, sarcoma, cartography, and so on. And then we performed experiments involving several conditions where people were alone trying to navigate the landscape, where people were in groups working together to navigate the landscape, or where people were in groups but also had some bots in there trying to help them be creative. And the bots worked because they could pass information from one region of the network to another by communicating with each other. So for example, bot number two could pass the word sky or car to bot number one. So here's a network example.
We introduce participants into the activity. The square markers represent automated entities. This automated entity engages in four interactions, while it possesses, I’m uncertain, six or seven interactions. Initially, the human participants begin by making guesses. They lack any foundation for their guesses in the early phase of the activity. They might suggest terms like sky, car, rabbit, rat, dog, cat, or desk. If I were to prompt you to select a noun to guess, you'd likely choose house, sun, moon, cat, or dog—typical simple nouns. Subsequently, they receive information about point values, indicating how closely related words like rat, dog, cat, desk, etc., are to the target word, which is sarcoma. Once the point values are communicated to them, this automated entity can evaluate the point values of the humans nearby and can either relay a random choice to this entity or the highest or lowest point value, establishing a sort of covert communication channel that shares information from one group attempting to solve the challenge to another. Now, let’s examine some illustrations to clarify this, as it can be somewhat complex. The participants engaged in this game for approximately 25 rounds. Here’s the cosine similarity related to a target noun, which is the term fratricide. There are 20,000 nouns in total. The fratricide is assigned 20,000 points. The other terms possess different point values. Here’s an individual participant, one who is guessing alone. Their initial guess is bit. They are informed how the term bit compares to fratricide. Then, for their next guess, they select birth. They experience a significant increase in cosine similarity, as it’s understandable that the term birth is more closely related to fratricide than bit. Next, they attempt money. That guess isn’t as effective. Then they try monkey, which is even less effective. They perform a sanity check and revert to birth, which boosts their score again. They consider baby as a guess, which isn’t too far off. However, baby pulls them even further from fratricide. They try lady and continue to navigate their guesses without nearing fratricide by the conclusion. In contrast, we have a different scenario with a collective of individuals but no automated entities. Here, the individuals can not only make their aspirations but also view the propositions of their peers and utilize those insights. It’s akin to collaboratively crafting the fishing rod. This individual initially guesses dog. But their subsequent guess becomes shield. You can observe that shield is nearer to fratricide than dog. This participant keeps making guesses, drawing insights from surrounding individuals, and ultimately progresses closer. They settle on the term foe, although soldier emerges as their most successful guess, achieving the highest point value throughout their journey.
So how did adding the bots matter? So this, again, shows the summary of the results. On the x-axis is the round. On the y-axis is the mean cosine similarity in this hyper-dimensional vector space, which is a measure of group performance. And here is a group of humans that are acting, guessing solo. So the humans that are guessing on their own don't do very well, right? They, they, you know, it's just chance whether they can somehow, or, you know, some innate ability in each individual human. They get a little bit better with time getting closer. All of the groups outperform the solo. And this is an old result that's known. A group of people is more creative than an otherwise similar in size set of solo practitioners. But what we find is, is that if we add the most similar bot, the bot that looked at its neighbors, [00:25:17] and found, what is their local consensus about here? My humans seem to be thinking that this is a good word. And then passes it to the bot at a distant part of the network. That bot substantially improved the performance of this group of people to make a discovery. You should be able to imagine how this could work in a group of engineers or any other knowledge workers. How, like, you can distribute knowledge in an efficient way, avoiding groupthink and fostering creativity, by designing bots that help the humans to help themselves. The bot here isn't, doesn't have a brain. It's not, it's self-suggesting ideas. It's just helping the humans to spread the ideas among themselves.
So simple AI agents with interpretable behavior can enhance the capacity for creative discovery in human groups by sharing ideas around which there is local consensus in one part of the group with people in a distant part of the group. And as a result, the group can perform better.
Currently, we have started exploring physical systems. We've incorporated robots, both humanoid and non-humanoid, [00:26:19] equipped with basic AI into face-to-face human groups, demonstrating how they can facilitate teamwork by aiding in overcoming tension or a lack of cooperation in interactions. A notable example from my past research with my previous graduate student, Maggie Traeger—who is now an assistant professor at Notre Dame—is an experiment we conducted. In this study, we invited three real participants into the lab alongside a humanoid robot. We created a simple game, akin to a train track layout game, that is operated on a tablet. This team of humans and the robot needed to arrange tracks from point A to point B, similar to a miniature Thomas the Tank Engine setup. We provided them with various track pieces on the tablet, like straight and curved sections. However, we cleverly designed the task so that there were not enough curves to successfully navigate from point A to point B, even though it appeared they had sufficient options. [00:27:21] As a result, they faced an unsolvable challenge. Each participant would take turns placing a track piece, working collectively to connect point A and point B on their tablets.
The process we undertook in this study involved manipulating interactions where participants engaged in 30 rounds of a game within a virtual setting, divided into 51 groups. We adjusted the way robots communicated, particularly focusing on instances where a robot acknowledged a mistake, indicating vulnerability, or opted to deliver dad jokes. The notion of dad jokes, widely recognized, seems to hold cross-cultural relevance, as anthropologists have observed similar humor among indigenous dads in areas like the Amazon. Kids there often respond to such jokes with disbelief regarding their fathers' humor. There's a theory suggesting that these jokes serve to toughen children, although that's a broader discussion. In our study, we had robots either share dad jokes or show vulnerability. Our primary interest lay in discovering whether changes in the manner of robot communication could influence not just interactions with robots but also how individuals relate to one another. Essentially, this presents an intriguing opportunity to modify social exchanges through the integration of artificial entities within combined human-machine systems.
So here's a little example of, oh, and I didn't tell you the results. So here are, here's when we have, so these, the thickness of these lines, we set video cameras up to monitor who's talking to whom and how much are they talking. And the thickness of these lines indicates how much person two is talking to person one and so on. So person one doesn't speak to the robot very much. That's a thin line. And when you had neutral robots, you get this pattern. But when you had the vulnerable robot, all these lines get thicker and they equalize. So we found that a vulnerable robot increased the equality of speech among the humans, increased the volume of speech among the humans, and actually, in separate results, [00:29:22] increased the satisfaction of the humans in that kind of environment.
And here's just one clip of two different rounds that illustrate the robot speaks in a neutral way first, in the passive voice, which doesn't much affect the human communication. the robot says, in a very Reagan-esque way, a mistake was made.
But in the next round, the robot says, I made a mistake. And you can just watch what happens here. Let's see if we can get this to work. (25 seconds pause) So across many, many dozens of runs, this is the kind of pattern we find. So a simple manipulation, simple manipulation in the robot speech pattern changes how the humans interact with each other. [00:30:25] And it doesn't, I'm assuming, doesn't take much to imagine how the whole way we design our chatbots and everything else might be affecting not just the Alexa example I gave you at the beginning, how we treat each other. The humans seem to trust each other more and to have more fun in this situation.
Currently, this is the final demonstration I intend to present, followed by a conclusion. In another experiment, we created an innovative cyber-physical platform aimed at examining the social and ethical implications of basic AI types. Considering the essence of collective action challenges, the integration of AI within human groups might unintentionally hinder existing positive social norms in humans, like cooperation and altruism, which we have developed over our evolutionary journey. Our species has benefited from extensive natural selection, enabling us to effectively address collective action dilemmas. The key inquiry is: if we assign some of that responsibility to machines, might we forfeit our ability to collaborate and resolve those challenges? [00:31:32] Will we become reliant on these machines, potentially diminishing our natural capacity for cooperation, coordination, creation, and so forth? In this project, in collaboration with Hiro Shirato at CMU, a former student of mine, we constructed a platform using two small Raspberry Pi-based mechanical vehicles, linking it to software that facilitates large-scale online experiments. Participants engaged from their homes, each assigned one of these vehicles, maneuvering them towards one another.
and we had them play the game of chicken. So in chicken, you know, like, whoever gets to the other side fastest wins. So you're incentivized to not yield to the other guy. But if each of you chooses not to yield, then you crash, and you both get the worst payoff. So what humans would do in this situation, if you're playing an iterated chicken game, is you would quickly learn to take turns. This time it's your turn to go straight through, I'll pull over and let you, but next time you pull over and let me go straight through. [00:32:35] If we're selfish, we just keep crashing into each other, like destroying each other time after time after time, or stupidly, we both swerve off, and neither gets the benefit of going straight. So in this situation over here, the yellow car decides to pull over, the blue car just proceeds unimpeded all the way to the other side.
We conducted a study involving 300 participants and 150 pairs, compensating them based on their speed in reaching the destination. Additionally, we incorporated some AI features. We implemented auto braking assistance, which activated a warning when approaching another vehicle, allowing the driver to decide whether to yield or maintain their position. We also included auto steering assistance, which caused the vehicle to veer away upon nearing another car. A minimal communication feature was added, enabling basic interactions like saying thank you. Our findings indicated that the auto braking assistance, where vehicles halt just before a collision, fostered human altruism, encouraging yielding, as illustrated by the yellow car. This AI enhancement facilitated cooperation among individuals in these scenarios. Furthermore, enabling communication among individuals allowed for mutual concessions under auto braking conditions. Conversely, the auto steering feature, which merely redirected the vehicle, inhibited reciprocity, promoting self-serving behavior. Individuals began to relinquish their moral responsibility, allowing the machine to repeatedly navigate without engagement. Consequently, the inherent ethical instincts of individuals were diminished in auto steering scenarios, while they were amplified in auto braking situations. This prompts reflection on the potential moral implications of AI programming choices on human behaviors.
The capacity of individuals to work together, share responsibilities, and show selflessness can diminish, ultimately resulting in poorer outcomes for everyone involved. [00:34:46] Indeed, in a nutshell, AI may foster a form of ethical complacency. (..) Here’s one last illustration of enduring transformation following exposure to AI in hybrid setups. It also exemplifies how the presence of AI can alter human-to-human interactions, even after the AI is no longer involved in these interactions. In 2016, DeepMind created AlphaGo, and that same year, this AI entity faced Lee Sedol, the exceptional world champion from Korea.
I viewed the match. I’m not proficient in Go, but I recognize it as an extraordinary game. My son plays Go. (.) I was genuinely cheering for Lee Sedol. In Korea, Lee Sedol is akin to a hero, similar to how we might feature notable athletes on Wheaties boxes, and so forth. His image appears on little noodle packets and cereal boxes. I find it remarkable that someone like me, a brainy nerd, can be recognized as significant in Korea, alright? He is very well-known there. When he steps up for the initial match, he comes off as overly confident. [00:35:47] I can see he’s too self-assured. He loses to the machine and subsequently apologizes to his supporters, saying he’s sorry. It's a best of five matches format. Then he plays a second match, loses again, and starts to take things seriously. He plays a third match, loses again, and he’s out of the competition.
As you observe the audience, the commentators are astonished by the machine’s Go-playing skills, executing strange yet beautiful moves. Some of these moves, they later discover, had historical significance as they have records dating back thousands of years of Go games played in the Chinese Imperial Court. They realize, oh wow, this move from the machine is a medieval tactic that we hadn't seen in a long time.
And then Lisa Dahl returns in the fourth game, securing a victory. (..) I was overwhelmed with emotion. I felt such joy. (..) Because he had the machine, you see, he had come back for my kind, and he had triumphed over the machine. In a heroic manner, it was beyond my comprehension the mental effort it took for Lisa Dahl to achieve that. [00:36:49] At that moment, I was filled with elation, and incredibly proud of him for persevering, even after the match had been lost.
Now, what's interesting is when Lisa Dahl was interviewed afterwards, Lisa Dahl said that his own game playing changed after the match. So once again, the AI is helping humans to help themselves, okay? He changes how he plays because of contact. And subsequent investigations by other scientists looked at professional Go players and looked at the median decision quality. There's some standard in Go for looking at how good a move is, and median novelty, like how unusual a move is. And they find that when AlphaGo defeats the human world champion in 2016, all around the world, Go players start making better moves that are more innovative, okay? So all the humans playing Go among each other have changed because of AlphaGo has been added. The AI helps the humans to help themselves. So we're continuing to build on our work to design and add simple bots to these and other situations involving social dilemmas and collective action problems. [00:37:55] We're looking at how bots can affect coordination, cooperation, communication, creativity, trust, navigation, sharing, and evacuation. And in our lab, we are not focused on super smart AI, like LLMs or AlphaGo, to replace human cognition, but rather on dumb AI to supplement human interaction. We're not trying to invent super smart AI to replace human cognition. We're inventing dumb AI to supplement human interaction. And our AI can afford to be dumb because the humans are smart. Our AI is like platinum added to an organic chemistry reaction. It's just a catalyst. All we need is the catalyst to help a group of people be better.
And of course, it's important to acknowledge that the reverse is also possible. Social AI can be used to harm groups of people. But our approach offers a number of other technical and conceptual advantages. First of all, these simple bots are intelligible and hence clearly illustrate broader powers and opportunities. Unlike LLMs, which are a black box and you don't know what it's doing, I can tell you exactly what our bot is doing. [00:38:58] It's noisy. It's brokering introductions. It's passing messages in this very specific way. And second, our controlled bot experiments can also provide insight into how human behavior could beneficially also change. In other words, I can take this from the lab and I can teach a group of humans to do what our bots did. In a way, you can't easily teach a group of humans just do what ChachiPT did here. We don't know what ChachiPT is doing, but you know what they're doing in our situation.
So I'd like to close, this is my last slide, with a metaphor. Consider these two objects. They are both made of carbon. If you take the carbon atoms and connect them one way, you get graphite, which is soft and dark. Take the same carbon atoms and connect them another way, you get diamond, which is hard and clear. And there are two key intellectual ideas here. First of all, these properties of softness and darkness and hardness and clearness are not properties of the carbon atoms. They are properties of the collection of carbon atoms. And second, which properties you get depends on how you connect the carbon atoms to each other. Take the same carbon atoms and connect them one way, you get one set of properties. [00:40:01] Connect them another way, you get a completely different set of properties. Similarly, the nature of our connections affects the properties of our social groups. It's the ties between people that can make the whole greater than the sum of its parts. New properties, such as cooperation and violence, innovation and productivity, trust and mistrust, truth and falsehood, wealth and poverty, health and happiness, can emerge and spread because of the connections, because of the ties between people, and not necessarily solely because of the people themselves. In fact, our experience of the world depends on the structure and function of the networks around us near and far. And our species evolved for this to be the case. And it should not surprise us that we will respond to AI in our midst. Thank you very much.
