[music] [music] People worry about AI for all kinds of reasons. It's polarizing. It spreads misinformation.
It's coming for our jobs. And those are all good reasons to be nervous. But what really keeps me up at night is something else.
something that feels more personal and perhaps even more concerning for the human experience itself. I worry that AI will make us boring. You, me, all of us.
That's because the more we outsource our decisions to AI, the more we surrender something deeply human. Our capacity to explore, to take risks, to stumble into the unknown, and sometimes surprise even ourselves. I worry without that without the potential for discovery, for serendipity, we risk becoming more shallow and unid-dimensional versions of who we are and who we could be.
Not just as humans, as individuals, but as humanity. Now, let me make this idea a bit more concrete and take you somewhere perhaps unexpected, but also totally relatable. Imagine walking into a BaskinRobins.
What you'll see is an illustrious assortment of 31 different ice cream flavors. There's chocolate chip, pistachio, almond, lemon sorbet, and many more. You're now facing a difficult choice.
Are you going to go with an old favorite chocolate maybe? Or are you going to take a risk on something funky and new like wild and reckless sherbet? And yes, it's a real flavor.
The choice you're facing is a classic human dilemma scientists call the exploitation exploration trade-off. Are you going to play it safe and capitalize on what you know you like? Or are you going to take a risk with the hope of finding something even better?
And our lives are full of these trade-offs. Stick with your favorite restaurant or try to fancy a new spot on the corner. Keep your usual haircut or ask for something new and edgy.
Stay in your stable job or finally try to become a pickle ball champion. We all differ in our appetite for exploitation versus exploration, but we're also all hardwired by evolution to strike a certain balance between the two. Because for our ancestors on the African savannah, the trade-off was actually pretty simple.
Stick too closely to what's safe, and you risk starvation when the grove runs dry. Strafe too far into the unknown, and you might get poisoned or eaten by lions. Luckily for you and me, the trade-offs we face today are far less about survival.
But our choices are still governed by this instinct to balance caution with curiosity. If all you ever did in life was play it safe and exploit, you'd never get disappointed, but you'd also never get a chance to advance and grow. If all you did instead was discover and explore, you'd collect endless experiences, but you'd also never get a chance to capitalize on your learnings.
So, it's really this balance between exploitation and exploration that fuels our growth, not just as a species, but also as individuals. But it's exactly where AI throws a wrench into the works. Because AI hates risk and that's not because of its algorithmic DNA.
It's because the systems we rely on to navigate our world from Spotify to Netflix to Chat GBT are overwhelmingly trained to optimize for exploitation or more specifically for short-term engagement and satisfaction. Did you click the link, watch the video, like the song? If yes, AI gets a clap on the shoulder.
If no, gets a slap on the hand. Risk, discovery, and exploration are simply not part of their programming. Think back to Baskin Robbins.
If 60% of people prefer the flavor pralines and cream, which is their bestselling flavor, then that's what the AI is going to recommend. It's not that AI lacks imagination. It lacks incentive.
Safe bets protect against disappointment, which in turn reduces customer churn. So instead of taking this risky gamble that helps you explore, companies tend to heir on the side of exploitation when training their AI systems. And to be clear, there's nothing inherently wrong with that.
I've spent the last 15 years as a computational social scientist working at the intersection of psychology, computer science, and business. And I know that rewarding AI in that way can be extremely valuable. Algorithms in a way only become so good at figuring out what you might enjoy or currently be looking for because they focus on exploitation.
And that's not just good for business. It's also good and sometimes even necessary for us as consumers. For me, the mere thought of having to choose between 31 different ice cream flavors makes me dizzy.
And yet, most of the choices that we face dayto-day are far more complex than that. Netflix, for example, has over 5,500 movies to choose from. Spotify over a 100red million songs.
So the truth is that without a little help from our exploitation loving AI friends, we simply don't stand a chance at navigating modern life. But I've also become increasingly concerned that this upside comes at a somewhat existential cost. And to give you a first flavor of what I mean by that, let's go back to BaskinRobins.
Before coming here, I ran a simple experiment. I asked ChachiBT to recommend one of their ice cream flavors. I did that a h 100 times each round pretending to be a new customer.
96 times it recommended one of their two most popular flavors, pralines and cream and mint chocolate chip, which I realize and I'm sure you'd agree is not exactly a very diverse set of choices. And it could mean that we soon find ourselves in a world where BaskinRobins offers only those two flavors. If no one ever picks the other 29, why bother offering them in the first place?
Now, this might sound trivial and not really existential at all, but that's only until you realize that this flattening of the human experience happens across every aspect of life. In studies with my collaborators and students, we've shown that when people use AI for guidance, their preferences become more normative and less diverse. Their creative output becomes less unique, and their choice of the most important scientists, athletes, and historical figures becomes the same as everyone else's.
So, in a nutshell, AI turns this infinite diversity of human opinions, beliefs, and preferences into statistically safeness. And that's only part of the problem because even when AI learns your quirks, say it figures out that you like nutty coconut better than pralines and cream, it will still play it safe within your preferences, meaning it is going to optimize for what you are most likely to like. I went back to Chachib for a second experiment and this time told it about my personal ice cream preferences.
I said that during my last 100 visits, I picked nutty coconut 70% of the time. The remaining 30% I split evenly among three of my other favorites. I then asked it to make the next 100 choices on my behalf.
And guess what it did? It picked nutty coconut every single time. Gone are my adventures into occasional chocolate fudge, cookie dough, or mango.
All I am at that point is literally a nutty coconut. But forget about ice cream. My students and I have repeatedly shown the same patterns when studying people's actual preferences.
AI narrows your taste. It flattens your personality and it scrubs away the edges that make you interesting and keep you dynamic. And the insidious part of all of this is that the impact of AI on human complexity is subtle.
You won't notice it happening overnight. Rather, it's a death by a thousand algorithmic recommendations. It's one slightly safer movie, one slightly more popular book, one slightly more mainstream vacation at a time.
With every decision, you outsource to AI. You become more narrow. Then AI learns from that shellar version of you and narrows its recommendations even more.
The New York Times reporter Kashmir Hill captured this dilemma perfectly in one of her articles. After outsourcing her decisions to AI for a week, she complained about its hidden agenda to turn her into a basic And the funny part is it's the word basic here that's the biggest insult because being basic means being unoriginal, unexceptional, and uninteresting. It's offensive not because it suggests failure, but because it implies mediocrity, an absence of complexity.
And who really wants that for themselves? Nobody. We all like the feeling of being unique, of being special, of being someone who doesn't easily fit into a box.
I, for one, don't want to become exactly like everybody else. And I also don't want to become a singular version of myself. So, what can we do to course correct and avoid destination basic We can't hit the reset button on AI and we shouldn't.
As I said before, these tools are incredibly valuable for navigating modern life and they make us better off in many ways. But we need to reclaim our ability to take risks and discover. We need to find a way to rebalance exploitation and exploration.
And ironically, AI could actually help us accomplish just that. But only if we ask it the right questions and reward it for the right actions. So instead of asking it to help us find something we like, we could start asking it to help us find something new, something we're likely to love even though we've never tried it before.
And by that, I don't necessarily mean you typing that question into ChachiBT yourself, although you should certainly try. What I mean is getting companies to harness AI superpower, its ability to detect patterns in vast amounts of human data for exploration purposes. Because AI has seen the entire universe of preferences, it knows not only what you currently like, but also what lies just beyond the boundaries of your typical preferences.
Which means that it can help you explore smartly to take curated risks when you want to. Now, here's what this could look like. Imagine a dial on your Netflix account or your Google search bar that lets you decide how far from your typical preferences you want to stray at any given point in time.
On a regular day, you probably keep the dial close to the spot-on setting because you want to get the most relevant hits right away. But then on other days, you might feel adventurous enough to push it a little closer to the me with a twist setting and ask for content that's a little outside of your comfort zone, but still relevant. And then on other days, for those rare moments when you feel like you're really ready to take on the world, you might push the dial all the way to the wild card setting and ask it to help you discover something entirely new.
I, for one, would love to have such a dial. It doesn't force me to leave my comfort zone. I can always leave it on the spot-on setting and benefit from the convenience that comes with it.
But it gives me the choice and the agency to break out of my little bubble whenever I want to or I feel like I have to. But here's the catch. For AI to toggle its superpower between exploitation and exploration, we need to incentivize it to do so.
merely asking it to be more creative or help you become more adventurous isn't going to do the trick. In most cases, it will still default to the tried and tested output because it craves that clap on the shoulder. So instead of punishing AI every time it takes a swing and misses, we need to start rewarding it for taking smart swings, for making bets that are a little bold, a little unusual perhaps, but still grounded in what it knows about us.
AI doesn't have to throw darts in the dark. It can take informed risks. It can help us optimize our exploration.
And when it does, we should treat it as a success, not a failure. Now, this isn't about desserts or playlists. This is about preserving what makes us uniquely, messily, and gloriously human.
It's your random passion for ET memorabilia, your weird detour into geocaching, or your stubborn preference for a stick shift car. It's about all the contradictions that AI can't quite explain. And the beauty of us being just a little weird once in a while.
And if we believe that human complexity is worth preserving, then the time to act is now. because we're at this inflection point where AI is no longer just recommending but starting to act on our behalf. It's choosing, not suggesting.
And that's where the stakes become really existential. So next time AI offers you pralines and cream, maybe say no and ask it to help you go wild and reckless instead. Thank you.
