Summary

This presentation, delivered at Google by neuroscientist, entrepreneur, and former Googler Anne-Laure Le Cunff, explores the concept of the "experimental mindset" as a framework for navigating uncertainty, fostering innovation, and avoiding burnout. Introduced by Alison Parrin, head of the Google School for Leaders, Le Cunff shares insights from her research at King's College London and her personal experiences to advocate for shifting our definition of success from linear goals to curiosity-driven learning.

The Trap of Linear Goals and the Productivity Paradox

Le Cunff begins with a personal anecdote from her tenure at Google. Driven by imposter syndrome, she overcommitted, overplanned, and routinely ignored early signs of physical exhaustion. This culminated in a medical emergency where she was diagnosed with a severe blood clot in her arm that threatened to travel to her lungs. Strikingly, her immediate reaction to the urgent need for surgery was to check her calendar to ensure it did not conflict with product launches.

This extreme response highlights a broader paradox facing modern knowledge workers: while hired to navigate complexity and think creatively, individuals frequently respond to uncertainty by over-compensating, over-planning, and seeking absolute control. This pathology stems from societal reliance on "linear goals"—neat, structured paths such as multi-year career plans, long-term mortgages, or rigid project launch calendars. While linear goals offer comfort, they fail in a highly nonlinear world subject to changing market trends, disruptive technologies, and global crises. When plans inevitably fail, individuals resort to self-blame, hiding their setbacks, and suffering from chronic stress.

The Scientific Definition of Success

In contrast to linear approaches, scientists define success not as reaching a pre-determined destination, but as learning something new. Within a laboratory, an unexpected result is met with interest rather than self-judgment.

Le Cunff explains that this mindset aligns with the brain's natural perception-action cycle:

• Perception: Gathering data from the environment.

• Prediction: Formulating a hypothesis.

• Action: Testing the prediction.

• Correction: Adapting predictions based on the outcome of the action.

While the human brain evolved to minimize uncertainty for survival, thriving in the modern world requires replacing this survival-driven anxiety with systematic curiosity. Research indicates that approaching challenges experimentally leads to faster problem-solving and significantly reduces anxiety.

The Framework of Tiny Experiments

To bring the scientific method into daily life, Le Cunff introduces the concept of Tiny Experiments. The framework consists of three main phases:

• Self-Anthropology (Observation): Observing oneself, thoughts, and environments objectively without preconceptions. This is powered by metacognition (thinking about thinking) to recognize habits, energy drains, and emotional reactions.

• The Pact (Hypothesis & Testing): Creating a simple, clear protocol to test a change. A "pact" requires only two ingredients:

• The Action: What is being tested.

• The Duration: A set number of trials or a predefined timeframe (e.g., 10 days). Deciding the duration in advance prevents confirmation bias and prevents abandoning the experiment prematurely.

• The Growth Loop (Reflection): Analyzing the collected data, drawing conclusions, and sharing results. By pairing action with reflection, individuals enter "growth loops" rather than endlessly repeating the same cycles of behavior.

Dismantling Cognitive Scripts

Our choices are often dictated by deeply embedded cultural "cognitive scripts." Le Cunff highlights three major scripts that limit individual growth:

• The Sequel Script: Making decisions based purely on past actions (e.g., feeling obligated to stay in a career path simply because it matches one's university degree).

• The Crowd-Pleaser Script: Making choices primarily to secure external validation, praise, or admiration from colleagues, family, or friends.

• The Epic Script: The belief that every effort must be massive, world-altering, and globally impactful. This leads to profound unhappiness for those who have not found a singular "passion," and causes an identity collapse when a large-scale project (such as a startup) fails.

Implementing the Mindset in Leadership and Teams

Le Cunff emphasizes that adopting an experimental approach is highly effective for teams. Leaders can build psychological safety and "social flow" by shifting from the expectation of having all the answers to facilitating collaborative discovery.

Key strategies for teams include:

• Leading with "I don't know": Demonstrating vulnerability and curiosity when facing unknown variables.

• Creating a Sandbox: Encouraging every team member to run their own tiny experiments and report back.

• Monthly Curiosity Circles: Gathering teams to share what they experimented with, what worked, and what unexpected outcomes or failures occurred.

• Redefining Success: Valuing people for the quality of their questions rather than the sheer volume of guaranteed answers.

Practical Applications and Q&A Insights

• Neurodivergence: The experimental mindset is highly suited for neurodivergent individuals (such as those with ADHD), as its nonlinear, curiosity-driven nature provides a flexible alternative to rigid, curriculum-like environments.

• Tracking Tools ("Plus, Minus, Next"): To track experiments, Le Cunff recommends a three-column framework:

• Plus: What went well.

• Minus: What did not go well.

• Next: What to adjust or implement in the next cycle.

• Limits on Scope: To prevent burnout and ensure clean data, individuals should run only one experiment at a time, or at most two experiments in entirely separate areas of life (e.g., one at work and one related to personal health).

Transcript

ALISON PARRIN: Welcome to Talks at Google. I'm Alison Parrin. I have the privilege of heading up the Google School for Leaders, which is Google's internal center of excellence for all things manager and leadership development.

Today, I am delighted to welcome Anne-Laure Le Cunff, who is an award-winning neuroscientist and entrepreneur. She founded Ness Labs, a platform supporting healthier ways to work and learn, and her newsletter is currently read by more than 100,000 knowledge workers. Her research at King's College London focuses on the neuroscience of lifelong learning and curiosity, which I'm sure we're all very fascinated to learn more about.

Her book, Tiny Experiments, is a transformative guide for living a more experimental life, turning uncertainty into curiosity, and carving a path of self-discovery. Previously, she worked at Google as an executive on digital health projects such as Wear OS apps and Google Fit. Her work has been featured in Wired, Forbes, and the Financial Times. Please join me in welcoming Anne-Laure to Google.

ANNE-LAURE LE CUNFF: Thank you. I'm so happy to be here today. I actually started my career right here in this office. So being back and being able to share some of my work and research since I left Google feels particularly special. So thank you for having me.

I'm going to start by sharing a photo I have never, ever shared publicly before.

As I mentioned, I used to work at Google. I started in London, and then I moved to San Francisco. And it was my dream job. I also was constantly worried that someone would figure out that they had made a hiring mistake, that I didn't belong there with all of these smart people. And as a result, I responded to that uncertainty by saying yes to absolutely everything and by desperately seeking a sense of control.

My task list looked something like that. I was also trying to timebox every single gap in my calendar. I was overplanning, overcommitting, and I was the "yes girl" in the office. Anything you needed me to do, I would say yes to. I was dangerously dancing with burnout. But I loved my job, I loved my team, and I loved our mission. So I kept on pushing through.

Until one day, I was in front of the mirror, brushing my teeth, getting ready for work, and I noticed that my entire arm had turned purple.

So I went to the Google infirmary in Mountain View. And the nurse had one look at my arm and said, "You need to go to the hospital right now." I went to the hospital. And there, the doctor said, "We need to perform surgery as quickly as possible. You have a blood clot in your arm that's threatening to travel to your lungs."

And what did I do in that moment in front of the doctors? I said, "One second. I need to check my calendar." And right there, in the doctor's office, I opened my calendar, and I proceeded to thoroughly check that the moment we would schedule that surgery would not conflict with any of the product launches I was working on.

As you can see, my arm is fine. It healed. We actually took that photo right after the surgery. But this moment stayed with me. It made me reconsider our entire relationship to work. It made me ask, why is it that so many of us push ourselves to the edge in the name of productivity? And it made me question our relationship to uncertainty in our personal and professional lives.

This is actually a really interesting paradox. As knowledge workers, we are hired for our capacity to solve problems, to think creatively, to deal with complexity. But somehow, when we're faced with uncertainty, we have this tendency to want to feel in control, to seek certainty. And because of that, we sometimes push ourselves to extremes—such as checking your calendar when someone is asking when you can schedule your surgery.

To understand it, we need to go back to the very definition of success. Success is something we all want. We praise it. We admire it. We want it. We also spend a lot of time trying to measure it. We have KPIs at Google, OKRs, performance reviews. We also very often measure our own success based on the success of others. It's as if we were all looking at a giant leaderboard, constantly asking who's doing better, bigger, faster work.

So what is this success that we're all so ardently chasing? If you open a dictionary, the most common definition of success you will find is something like this: reaching a desired outcome.

Now, please indulge me while I break down this definition and we try to understand what is this definition we've all agreed on as a society.

Reaching—there's the idea of movement, progressing towards something. Towards what? Something we desire, something we feel like is positive. And that thing is an outcome, which implies that in order to be successful, you need to reach a specific destination.

This might seem like the only obvious definition of success, but it's not. This definition of success is actually based on something called linear goals. Linear goals are based on the idea that in order to be successful, you need to have a clear vision and a clear plan. And if you start looking around, you'll notice that those linear goals are everywhere in our life and in our work.

We have four-year university degrees, followed by a five-year career plan, followed by a 30-year mortgage. And even in the way we manage our work on a daily basis, we might say, "Here are the features we're going to launch in the fall. And here are our sales targets. And here are the marketing activities that we'll put into place in order to reach those linear goals."

And it feels really good to have this sense of certainty. But there's only one problem with linear goals: it's that we don't live in a linear world. We live in a nonlinear world.

You have market trends that keep on shifting, new technologies that can disrupt your industry, and, as we've seen recently, global events that can change everything overnight. And so things rarely go to plan. Instead of going from point A to point B in a very neat way, we find ourselves navigating this complex web of twists and turns with unpredictability at each crossroads.

And what do we do when we can't achieve our goals? We blame ourselves. And very often, we might even want to hide our failures from others. In today's world, in this nonlinear world, clinging to linear goals can only lead to frustration, to overwhelm, and very often to burnout.

So, you know who has a completely different definition of success? Scientists.

For a scientist, success is not reaching a specific destination. Success is learning something new. Whatever the outcome, whatever the results, they're able to look at it without self-blame or self-judgment.

And today, I want to convince you to start treating your work and maybe your entire life like a laboratory. I want to convince you that being curious is much more powerful than feeling certain. I want you to start imagining what it would look like if we approached any challenge as an opportunity for experimentation. And to do that, we're going to study how to develop an experimental mindset.

So the first thing is, how do scientists react when they get an unexpected result? When a scientist doesn't get what they expected, they don't say, "Shame, shame, shame, I'm such a bad scientist." No, they look at it and they ask themselves, "Huh, what's going on here? What can we learn from this?" And this is because they understand that we need failure to learn. Failure is an inherent part of learning.

What's interesting is that this kind of thinking, knowing that failure is a part of learning, is actually aligned with the way your brain works—or, should I say, the way your brain would like to work if you didn't force it to follow linear goals.

The way your brain works is based on something neuroscientists call the perception-action cycle. It's fairly simple. First, you're going to perceive some information in the environment, some data. Based on that data, your brain is going to formulate a hypothesis, a prediction. Sometimes that prediction is correct, and it's great. Sometimes the prediction is wrong, and that's okay, too. As long as you're still alive and you're not dead because of that wrong prediction, your brain is going to use that new data, that new information, and make a new prediction. And this is how you learn through experimentation.

Where it gets a little bit more complicated is that your brain is also optimized for survival. So it's trying to reduce uncertainty as quickly as possible. And that makes sense from an evolutionary perspective. If you think back on our ancestors and the environment in which they were evolving, the more information you had—whether it's, "Where are the resources?" or "What's that weird noise in the bushes over there?"—the more you knew, the more certainty you had, the more likely you were to survive.

But I think we all agree that whether it comes to our work, our relationships, or our health, us modern humans want more than just surviving. We want to thrive. And in order to do this, we need to replace this desperate need for certainty with curiosity instead.

You're still using your perception-action cycle. But instead of trying to resolve uncertainty as quickly as possible, you're using that uncertainty as an opportunity to learn and to grow. What's amazing is that there is research showing that when you approach challenges in this curious way, in a more experimental way, not only are you going to find solutions faster, but you're also going to experience less anxiety and stress in the process, which is pretty neat.

The reason why scientists are so good at this, at having this experimental mindset, is not because they're smarter than all of us. It's because they've been trained to do so.

For anyone who has studied science at school, you probably remember this experimental cycle. And again, it's quite simple:

• You start with observation, where you ask, "What is the current situation?"

• Then, based on that current situation, you formulate a hypothesis. You ask, "What could be different?"

• Then you start data collection. You test that hypothesis.

• And finally, you analyze the data. Based on those results, you update your observations so you can design your next experiment.

This is the experimental cycle. It can only work when you pair action with reflection. You need to do something, look at the result, and then change the way you behave based on that information that you just collected.

When you use them all together, they form an operating system for how to develop an experimental mindset:

• First, knowing that failure is an inherent part of learning;

• Believing that curiosity always, always beats certainty;

• And always pairing action with reflection, reflection with action.

When you use this operating system, not only are you going to be able to navigate uncertainty in a smoother way, but again, you're going to be able to keep your sanity in the process and not feel as stressed and anxious.

The nice thing about this experimental cycle that I just showed you is that you don't need a lab to use it. You can actually run your own tiny experiments for any challenge that you're facing in your life and in your work. What I'm about to show you is, in essence, a very simple way to take the scientific method out of the lab and to apply it to any area of life and work, so you can go from that free-floating anxiety to systematic curiosity.

I know that the kind of people who would be interested in this talk are the kind of people who are probably problem solvers, who like getting things done, and who are quite creative. I know it's exciting to execute on something, but not so fast. When you want to design an experiment, it always, always starts with observation.

I like to call this self-anthropology. Just like an anthropologist goes and studies a new culture with no preconceptions whatsoever, you can actually study the way you think, the way you live, the way you work, and pretend that you don't know anything about the way things are done. Really take notes and ask yourself, "Why are we doing things the way we are?"

This is really an exercise in paying attention—paying attention to how things are so you can start imagining how they could be. And this is what will help you plant the seed of a hypothesis. You start with observation, which then allows you to imagine something that you might want to try.

Then you're ready to create your mini protocol for experimentation. When you want to experiment, you don't have to have a full, complicated experiment like a scientist would in the lab. You only need two ingredients:

• You need to know what you're going to test.

• You need to know the number of trials (the duration).

You need to know the action and the duration. And that's it. This is your protocol for experimentation.

I call this a "pact" because it's a commitment to curiosity. It's a commitment to collecting the data and withholding judgment until you have the results. You say, "I am going to commit to trying this thing for this duration."

It's very important to commit to the duration before you get started. First, you need several trials to know if something is working or not. If not, it might be just a coincidence. Second, if you don't commit to your duration in advance, you might be tempted to stop the experiment in the middle if you're not seeing what you want to see. This is why scientists decide the number of trials in advance. This way, you avoid confirmation bias—finding the result that you actually want to find—and you withhold judgment until the end, when you can actually look at it and analyze all of the data together.

I want to show you how flexible this approach is. You can use tiny experiments for literally anything. You can experiment with acquiring new skills, trying new tools, doing new research, or connecting with new people. Here, I've focused on work-related examples, but you can actually use tiny experiments in literally anything. I've seen people run tiny experiments to experiment with their health, meditation, creative hobbies, and even with dating. So it works with literally anything.

The last step, once you're done collecting your data, is to reflect on the results. This is the moment where you take some notes, where you might want to discuss it with other people, and where you share with others what you learned. This is the last part of the experimental cycle, and this part is extremely important because this is what allows you to learn from the experiment and to close this loop. That way, you can implement whatever you learned into the next cycle of experimentation.

This is what allows you to not just keep going in circles, but to really grow through the cycles. This is what people call "growth loops"—where you grow through each loop that you close, implementing the data into the next experiment.

How do you embody this experimental mindset? How can you actually learn and lead like a scientist?

The great thing about tiny experiments is that you actually don't need to get any kind of buy-in from anyone. You just need to notice that maybe something might be worth trying, and that's enough to design a tiny experiment.

That being said, we can actually grow and learn better and much faster when we experiment together. To do this, we need to reimagine some of the ways that we envision leadership. Whether you're leading a project or a team, you might be putting a lot of pressure on yourself to look like you know where you're going, to look like you're the expert, and to look like you have all of the answers.

But one of the most powerful things that you can do as a leader when faced with uncertainty is saying, "I don't know, but let's figure it out together." This is how you open a space for experimentation and for learning in public, where everybody is learning together, including from our failures.

Even better, you can encourage people around you—actively encourage them—to design their own tiny experiments. This way, you can unlock "social flow," where that information is flowing between team members and everybody can grow together.

For instance, what might it look like if, with your team, you were hosting a monthly curiosity circle where everybody would share their experiments, what worked, and what didn't? This is a great way to learn together and to create that safe space for experimentation as a team.

For all of this to work, though, we need to redefine success, not as a fixed destination that is based on linear goals, but as something that we learn together. Success is learning something new.

Ultimately, this shift in mindset is all about defaulting to curiosity. It's about learning to fall in love again with problems. It's about letting go of the fear of failure, the imposter syndrome, and analysis paralysis. It's about internalizing the belief that if you approach it with curiosity, any challenge can be an opportunity for growth and discovery.

Tiny experiments can lead to big changes. Imagine a culture where it's completely normal to walk around and to ask people, "What have you been experimenting with? What did you learn? What was your latest failure?" Not only would this be the kind of culture where there's more space for innovation and imagination, but that would also be the kind of culture where we value people not based on the quantity of answers they provide, but based on the quality of the questions they ask.

And this—this is the power of an experimental mindset. Thank you.

ALISON PARRIN: Thank you for such a bold invitation to consider how we are operating and what we might need from ourselves in a world that is consumed with certainty, knowing, and specific goals. I see such possibility and opportunity in the ideas that you share. I'm really excited that we have this opportunity to be able to discuss them together.

So let's start with this notion of certainty. You talk about how we are wired for certainty, and I think that's one of the things that we see. What are ways in which we can become more comfortable with uncertainty? Experimentation is obviously one route, but how should we be thinking about that and just embracing that discomfort?

ANNE-LAURE LE CUNFF: I think the first step is to acknowledge the fact that it's completely normal to feel anxiety when we're faced with uncertainty. Again, that's what our brains are designed for: to reduce that uncertainty. Whenever we're faced with a situation where we're not quite sure what's going on, what the threats or risks are, or who the other players are, our brain wants to reduce that uncertainty as quickly as possible.

So I think there is sometimes a lot of self-blame around uncertainty, where you feel like, "How come everybody looks so comfortable and I'm so scared?" That's the first step: knowing that that's normal, and it's probably the case that other people around you are just a bit better at hiding it. We're all feeling scared when we're uncertain. That's step number one.

And number two—this is why the book is called Tiny Experiments. We don't have to necessarily go for something really big and scary straight away. We can start with something very small. This is how you start building that muscle of playing with uncertainty and having a relationship with it that is similar to a scientist's. When they see something they don't understand, they actually light up. They feel like, "Ooh, juicy. There's something interesting here."

You start tiny. You start by looking at little things you don't understand. The more you do this and the more you experiment, the more you're going to find yourself in situations where you're out of your depth, but somehow, you feel like that's exciting.

ALISON PARRIN: One of the things you talk about there is the importance of observation and seeing different things and how you're reacting to them. How can we improve our skill of observation and do it more frequently?

ANNE-LAURE LE CUNFF: There's a uniquely human capability called metacognition. Scientists love jargon, but it really just means "thinking about thinking." We know that most mammals are able to think. Anyone who has a pet knows that—you look at a cat or a dog, and you know they can think, right? But as humans, we're able to observe our own thoughts, our own emotions, and our own behaviors.

I think this is a great way to start with observation: by turning that eye, that attention, towards yourself. We're usually better at observing the outside world. We're happy to observe what's going on, take a few notes, and share them with the team. But it's a bit more uncomfortable to just observe how we're feeling—the tension, the uncertainty, and the anxiety that we can have when we're navigating challenging moments at work and in our personal lives.

So something very simple that anyone can do is just taking a little bit of time every day to write a few notes and just observe: How was today? How did you feel? Not just the external measures of how things went, but how did it feel internally? By practicing doing this, you'll become a lot better at naturally observing how things are around you before making any decision.

ALISON PARRIN: So journaling and reflection—is that best done individually, or can I do it with other people? Is there a better way?

ANNE-LAURE LE CUNFF: The better way is the way you actually do. Journaling is great; there's so much research showing how good it is for your mental health and your creativity. But the fact is, lots of people don't like it. So that's why I tell people, if that's not working for you, there are lots of other ways to engage with active observation.

If, for you, it's finding either a friend or a colleague you feel quite close to and saying, "Hey, once every couple of weeks, let's grab coffee together—and this is just to share how we're doing. That's it." You can talk about work, about mental health, or about creative projects you're exploring. When you verbalize what is going on in your mind and with your emotions, it really forces you to understand and articulate them. Whether you do this through writing in a journal or through talking with someone, it doesn't really matter, as long as you do it.

ALISON PARRIN: Got it. I can imagine people wondering, in a world that already feels very full and somewhat overwhelming, "I just don't have time for that. I can't find the time, and I'm not really sure why that's going to be beneficial." What would you say to them in that context?

ANNE-LAURE LE CUNFF: I would say, just experiment with it. That's why my book is actually not that prescriptive in terms of how you implement these things, because I think it looks different for everyone. The idea here is to experiment with different ways for you to pay attention to how you feel, your productivity, how you work, communicate, lead, and relate to other people, and then run tiny experiments so you can see what works and what doesn't. You can adjust your approach and adapt instead of sticking to the same rigid approach over and over again.

For some people, that might look like taking only two minutes a week to reflect on the important things that came up in the past week. So I don't believe that nobody has time for it. We also know that we spend a lot of time doing other things that are not so good for us. If you took five minutes out of scrolling on your phone and used that for self-reflection, you would probably benefit a lot from it. It's rarely a matter of not having enough time; it's more a matter of not having found the right way for you to do this. You can't find that way just by reading a book. You actually need to experiment and see what works for you.

ALISON PARRIN: The key is the word "tiny"—that it really can be small.

ANNE-LAURE LE CUNFF: Yes, at least when you start. It can grow bigger if something you like actually works really well for you.

An example of a tiny experiment that I actually run myself: I used to be terrified of public speaking. I'm talking terrified as in stomach cramps and nightmares for weeks before I had any kind of presentation. So I asked myself, "What is the tiniest experiment I could run around this?"

I decided that, for the next 10 days, I was going to record myself with my phone for one minute, and post it on Instagram unedited. One minute. It was absolutely terrifying. But after a few days, I could already feel like opening my phone and starting to record myself was less and less scary.

After I finished that experiment—the set duration for this pact—I said, "Actually, I think I kind of liked it towards the end." So I asked, "What is a slightly more ambitious version of this experiment?" For this next phase, I decided that every month, I was going to find an online workshop that I could present. I wasn't ready to go on stage in person yet, but from the comfort of my home in my pajamas, I could do this.

Once I completed that duration, I moved to the experiment I'm currently doing: once a quarter, I need to find something quite big and scary where I need to be on stage in person. I'm already starting to feel a little bit less anxious.

So you can keep them tiny, and you should certainly start tiny. But if, in the process of experimenting, you discover that something is quite interesting and you want to grow through this and experiment more, you can also make them a little bit bigger.

ALISON PARRIN: Got it. I appreciate the fact that you've experimented in that way, because by doing that, you've shared with us the gift of your knowledge. We would never have that had you not been able to do that. So thank you.

One of the things I find fascinating when I looked at how you were describing experiments was the fact that there isn't actually a hypothesis there in the visual. It says, "I'm going to do X by X or for X period of time," but there isn't a statement of what you expect to learn. Can you talk more about that lack of the predefined thing that you might learn, and what that opens up?

ANNE-LAURE LE CUNFF: Yes. I didn't want to put the entire book in the presentation, so I skipped over that part a little bit. But there is a hypothesis, just not exactly as a scientist would write one in a paper. When you run an experiment, the hypothesis is usually along the lines of, "This is going to work," or "This is not going to work."

For example, for me, I was very scared of public speaking, and I had the hypothesis that maybe starting by recording those little videos might help. But you can also run experiments with the hypothesis that something is not going to work.

A personal example: a great way to find experiments is when you hear yourself saying something that sounds like you have a fixed mindset. I was talking with someone about meditation, and I said, "I'm so bad at meditation. It doesn't work for me. I tried." We all know those apps with the 10-session onboarding for 10 days. I had never managed to get past day three. That's how bad I was.

When I heard myself say this, I thought, "Oh, wait a minute, that's interesting—fixed mindset here." So how could I be more experimental with this? I designed an experiment, and I started fully convinced that it would not work. But I wanted to experiment anyway, collect the data, and see the result.

I committed to meditating every morning for 15 days. I decided to actually run this experiment in public. I kept a public Google Doc where, every day after I meditated, I wrote some notes and shared it online. I had a lot of people leave comments and give me advice. When I wrote, "Why is it itchy everywhere? Why can't I stay still?" people replied, "That's completely normal. Here are some techniques."

Not only did I complete the experiment, but I actually enjoyed it. I ended up being wrong. Now, I don't meditate every day—it's not that kind of miraculous story—but it's part of my toolkit now. If I feel particularly anxious, I'll sit for 15 minutes and meditate, which was something I could not have imagined before. So you can absolutely have a hypothesis, even if it is simply, "I think this is not going to work, but I still want to try it."

ALISON PARRIN: In that example, the idea of public accountability sounds like it was powerful. What role does accountability play with experimentation?

ANNE-LAURE LE CUNFF: You can run your experiments on your own; you don't have to share them with anyone. But it can really help to add this layer of learning in public, especially if it's something where you have quite a bit of personal resistance.

Maybe you've tried it before. Maybe it was a habit you tried to build in the past, and you couldn't do it. With habits, I find it completely crazy that we pick a new habit and say, "I'm going to commit to this for the rest of my life," when we've never even tried it before. You should run a tiny experiment first to see if it works, and if it does, then turn it into a habit.

Learning in public and having that accountability can be helpful to actually stick to it and collect the data. I highly recommend that if you're running an experiment where you feel like you'll be tempted to quit in the middle, do it in public. That's going to be helpful.

Second, in the spirit of shared knowledge and generosity, if it's an experiment where your friends, family, or team might benefit, you can share it with them. What's really important is to remember that it's not just about sharing the experiments that worked, but also the ones where you got an unexpected result. There is a lot of value in saying, "Hey, I tried this thing. It doesn't work. Don't do it." You're saving people a lot of time and energy by doing this, and it can be an amazing contribution. So learning in public is completely optional, but can be really helpful.

ALISON PARRIN: As a leader of a team, how would you encourage me to try this with my team?

ANNE-LAURE LE CUNFF: I would ask each person on your team to design a tiny experiment. You could pick a theme, a product, or a challenge that you're facing as a team and say, "For the next month, let's all run an experiment. And let's all report back at the end of the month."

Everybody can share what worked and what didn't. When you have this kind of scaffolding for experimentation, you're creating a sandbox—a playground where it's okay, and even encouraged, to share unexpected results. Instead of saying, "This is success, this is where we must go," you start from a research question or a hypothesis: "We think this might work, but we don't know. Let's try it, collect the data, regroup, and learn from each other." This is a very simple and tactical way to do it.

At a more strategic level, it's about leading by example. This means being okay with saying, "I don't know," but matching it with, "Let's figure it out together. Is there an experiment we can design to find the answer?" That is how leaders can truly encourage their teams to develop an experimental mindset.

ALISON PARRIN: Yes, and I think that comes along with a lot of courage, in terms of being able to say, "You know what? I don't know." That's certainly one of the things we work with our leaders on here.

One of the topics I found fascinating in your book was the idea of cognitive scripts. I'm wondering if you might be able to share a few thoughts around those.

ANNE-LAURE LE CUNFF: This is a fascinating concept based on an elegant study from 1979. Researchers basically asked people, "If you are put in this specific situation, how do you act? What do you do?" What they found is that most people, when placed in the same scenario, end up acting in exactly the same way.

This is highly useful in many scenarios. For example, if you go to the doctor, you know you are supposed to wait in the waiting room until they call your name, then go into the office, and then they check what is wrong. If the doctor comes out of their office into the waiting room and asks you to undress in front of everybody, you would feel extremely uncomfortable. That's because they've gone off-script. There's a script we've all agreed on, and the doctor is not following it.

So we have all of these scripts that are very useful for functioning as a society—going to a doctor, going to a restaurant, and so on. The problem is that scientists discovered we also follow these cognitive scripts in many other areas of our lives: how we choose our jobs, our careers, how we dress, the way we talk, and the subjects we study.

In the book, I share three big cognitive scripts—or rather, buckets of scripts—that are useful to notice in your life or work:

• The Sequel Script: This is when you make decisions based entirely on the decisions you made in the past. You feel like whatever you decide to do today needs to make sense based on what you did yesterday. This is why a lot of people, when they finish university, only look at jobs that align with their studies instead of considering other options. It's also why we rewrite our CVs when applying for new jobs to make it look like we had a neat, intentional narrative all along.

• The Crowd-Pleaser Script: This is when you make decisions based on what you think will make the people around you happy—whether that is your team, friends, family, or spouse. We often limit the scope of our decisions because we only go for things that will be praised, admired, or recognized by others.

• The Epic Script: Inspired by Hollywood, this is the idea that whatever you do, it must be massive, highly impactful, and save the world—and anything less than that is a failure. This one is particularly insidious because, as a society, we've decided this is a script we should all follow: Follow your passion, follow your dream, change the world. Because of this, a lot of people feel miserable because they haven't found their passion yet, wondering why everyone else has figured it out. Another problem is that people put all of their eggs in one basket. When that one thing doesn't work out, their entire sense of self-worth and identity collapses. We see this a lot with startup founders when their company fails, leaving them depressed for months or years.

I highly encourage everyone to think about these scripts and identify different areas where you might be following them at a subconscious level. Use your observation skills and ask, "Is there a way I could do things a little bit differently? Is there a way I could experiment with an approach that is slightly off-script?"

ALISON PARRIN: Yes, I think we have a lot of homework to do when we get back, in terms of thinking about how some of these patterns show up. I can see a lot of those patterns in my own stories.

In a moment, we're going to move to Q&A. If anybody has any questions, please begin to line up behind the microphone. While you're doing that, I will ask you one final question. I am excited to leave here this afternoon and begin a tiny experiment. What advice do you have?

ANNE-LAURE LE CUNFF: I'll go back to starting with observation. I recommend doing a 24-hour exercise in self-anthropology.

Choose a day during your week that is a pretty typical day—so don't do this on a Saturday when you are going to a festival. Do it on a normal workday. Just like an anthropologist, start taking little notes throughout the day, in between meetings or tasks, and ask yourself: What is giving me energy? What is draining my energy? When do I feel particularly curious and excited?

You'll very quickly notice patterns. Maybe you just finished a meeting and were particularly excited, wanting to spend more time on that topic. Equally, maybe you just had a conversation with someone and all you wanted to do was crawl into bed and disappear. Make little notes of these moments—the good, the bad, and the challenging.

When you notice these patterns, they can become the seed of a hypothesis. How can I do things differently? Maybe you've been running your meetings exactly the same way forever without ever questioning why. Maybe you've been working on the exact same project for years without questioning if there might be another project you'd enjoy more. Maybe you can experiment with your time management, calendar, or productivity.

So that's my advice: pick one day, do 24 hours of self-anthropology and observation, and then choose one tiny experiment with one action and one duration. Keep it tiny, don't go for something massive at first, and see what you learn.

ALISON PARRIN: I love that. Thank you. We have our first question from the audience, please go ahead.

CHAUNCEY: Hi, I'm Chauncey. I heard you were an APMM (Associate Product Marketing Manager) when you were here previously. I was too, and graduated recently, so it's really cool to see how far you've come and the great work you're doing. I'm actually going through a big life stage at the moment where I think this book is literally the Holy Grail; my mind is blown. I had one question around neurodivergence. How do you see these experiments affecting those who might move through the world with their brains wired a little bit differently?

ANNE-LAURE LE CUNFF: This experimental mindset is actually perfect for neurodivergent people. My job at King's College London is actually based at the ADHD Research Lab. Although this book is not explicitly about neurodiversity, that was always on my mind while writing it.

What you find with neurodivergent people is that they often have a more nonlinear way of thinking. It can feel incredibly constraining and uncomfortable for them to follow rigid, step-by-step curriculum-like approaches. The experimental framework is perfect because it starts with curiosity—which is typically very high in neurodivergent individuals—and leverages it to explore, experiment, and organically discover what works and what doesn't.

CHAUNCEY: Amazing. Thank you so much.

ANNE-LAURE LE CUNFF: Thank you.

ALISON PARRIN: We have a couple of questions coming in on Dory, but we'll take one more in the room first.

SPEAKER 1: Hey, Anne-Laure, lovely to see you. I used to work with Anne-Laure many, many moons ago, so it's wonderful to see you again. I was actually thinking of a very similar question to the one just asked, so I'll ask a different one instead. I'd love to know, what have been some of your own personal favorite tiny experiments, and have there been any surprising learnings along the way?

ANNE-LAURE LE CUNFF: I love the meditation one because I love the experiments where I start out convinced that it's not going to work, and then I'm proven wrong. As a scientist, both in and out of the lab, being proven wrong is the best feeling.

Another one I did recently was very simple but extremely good for me. While I was on my tour for this book, I had to record a lot of podcasts. At some point, I realized that on some days, I was indoors from 8:00 AM to 8:00 PM with back-to-back meetings and zero breaks. So I thought, "Maybe taking some walks will help."

I made a commitment—which is key to the tiny experiment format—and said, "I'm going to take a 20-minute walk every day for the next 20 days, and see if it helps." I did that for the last 20 days of my book tour. This time, I was not proven wrong; it really, really helped my mental health.

This just goes to show you that experiments don't need to be complicated or groundbreaking. You don't need to reinvent the wheel; it can be something very simple. Thank you.

ALISON PARRIN: One of the popular questions on Dory: "Do you recommend specific practical tools, like journaling, specific apps, or mental models, for tracking, managing, and reflecting on experiments?"

ANNE-LAURE LE CUNFF: There is a very simple tool featured in the book that is helpful for tracking, reflecting, and deciding what to implement in your next cycle. If I had named it while I was still working at Google, the name might be snazzier, but I came up with it on my own: it's called Plus, Minus, Next.

It features three columns:

• Plus: You write down everything that went well.

• Minus: You write down everything that didn't go so well.

• Next: (with a little arrow) What you want to tweak, adjust, or implement in your next cycle based on what you just learned.

I usually use this as a weekly review for my experiments on Sunday evenings or Monday mornings. If you are running a short, daily, highly intense experiment, you can do it every day since it only takes a few bullet points per column.

The nice thing is that if you like to do some form of annual review, you can look back at all of these templates at the end of the year and see all the experiments you've run over the past 12 months. It's a wonderful tool for reflection.

ALISON PARRIN: Love it. Thank you. Next question in the room, please.

SPEAKER 2: Hi, thank you. Joffrey here. My question is around the current climate. I think everyone knows that the company and a lot of departments have gone through a lot of changes recently, and anxiety around job security is at a certain level. Do you have any tips on how to influence leadership or the culture in general to make it more "failure-friendly"?

I feel like the "scripting" to avoid failure is stronger than ever right now. Sometimes we are subtly encouraged not to report on certain numbers if they don't look good, and to shape the narrative to look a little better. It would be wonderful to influence the culture so we can actually learn from each other and treat failures as a good thing.

ANNE-LAURE LE CUNFF: Yes, absolutely. This is why it is incredibly helpful to frame any project that has a lot of uncertainty as an "experiment."

The issue arises when we have a highly uncertain project and we link it directly to a rigid, linear goal or a specific destination. When it doesn't work out and we don't reach that destination, the exact behavior you described happens: we try to construct a post-hoc narrative explaining why it failed, why nobody is to blame, and how we'll do things differently next time. But we aren't truly learning because we're just trying to hide the failure and make it look like a success. The incentives get skewed, and the process is no longer designed for learning.

If you start from the very beginning by explicitly stating, "This is just an experiment. Here are the parameters. We think this is going to work, but we aren't entirely sure. We will run it for this set duration, and then report back on exactly what worked and what didn't," you completely transform the definition of success. It shifts from a binary outcome—where you either succeeded or you failed and have to hide it—to a collaborative process where the sole goal is to learn something together.

SPEAKER 2: Thank you.

ALISON PARRIN: Next question, please.

SPEAKER 3: Hi, Anne-Laure. My question is about the complexity of internalizing this experimental mindset. We all have different baselines—some have a genetic propensity to be more anxious or crave certainty, childhood experiences can shape our associations with risk, and then we have the prefrontal cortex telling us we can reason through this and change.

How malleable is this trait? I know practice is essential, but do we eventually have to accept our personal limits with uncertainty, or can we keep practicing and eventually become fully, comfortably experimental?

ANNE-LAURE LE CUNFF: This is a great question, because a common misconception about the experimental mindset is that it's designed to completely eliminate uncertainty and anxiety. It isn't. Anxiety is a completely natural evolutionary reaction, and it is incredibly difficult to get rid of entirely.

What you want to do instead is notice it, accept it, recognize that it is perfectly normal, and then design an experiment around the challenge anyway. By converting your anxiety from paralysis into active experimentation, you naturally reduce the uncertainty. You are never going to eliminate the fear of uncertainty completely—that's simply not possible. But experimenting gives you a sense of agency. You can say, "This is scary, and I don't know what's going to happen, but I have agency. I can experiment and discover my own answers."

ALISON PARRIN: Love that. Thank you. I think we have time for one last question.

SPEAKER 4: Hi, Anne-Laure, thank you. I was reading your chapter, "A Deeper Sense of Time," right before the talk, and I completely agree with your views on how we take productivity hacks too far.

I wonder, could designing tiny experiments and trying to get good at things in this structured way be interpreted as just another productivity hack? Is there an upper limit to this? Is it possible to do too many experiments and inadvertently push yourself back into burnout? How do those two ideas coexist in your head—not obsessing over productivity, yet constantly designing experiments?

ANNE-LAURE LE CUNFF: Yes, that's a very fair question. First, I don't view tiny experiments as a productivity hack in the traditional sense, but I do think they can help you discover ways to be productive without sacrificing your mental health. It is about questioning the way you work to find a gentler approach that yields the same result.

To your second point: yes, there is absolutely such a thing as running too many experiments. I highly recommend running only one experiment at a time for two main reasons:

First, you want to actually complete the experiment. If you are trying to run five, six, or seven experiments simultaneously, it is highly unlikely you will have the bandwidth to stick to them and collect clean data. It's much better to focus on just one.

Second, if you are changing multiple variables in your life and work at the same time, it becomes impossible to isolate which experiment is actually having a positive impact and which one isn't working.

The only exception is if you really want to run two experiments at once, ensure they are in completely separate areas of your life—for example, one work-related experiment regarding calendar management, and one personal experiment regarding your diet or health. But if you can, stick to just one experiment at a time.

SPEAKER 4: Thank you.

ALISON PARRIN: Well, thank you. Thank you for joining us. I am really excited, as I said, to go away and try this. I will invoke the power of observation over the next 24 hours, and I look forward to seeing what we learn. Thank you for the invitation to all of us to think about how to experiment more and become more comfortable with uncertainty. Thank you.

ANNE-LAURE LE CUNFF: Thank you so much for having me.

Summary

The Drive to Build Artificial General Intelligence (AGI)

Demis Hassabis dedicated his life to artificial intelligence over 30 years ago, driven by a childhood fascination with reality, consciousness, and the universe's biggest questions. Viewing AI as the ultimate tool for scientific discovery, he believes that building AGI is the key to understanding the profound mysteries of the human mind. By creating an intelligent system and comparing it to human cognition, scientists will finally have a reference point to isolate and study concepts like consciousness and true creativity.

Revolutionizing Biology and Drug Discovery

Through DeepMind and Isomorphic Labs, Hassabis is leveraging AI to drastically accelerate medical breakthroughs:

• Beyond AlphaFold: While predicting stable protein structures was a monumental first step, the focus has shifted to biochemistry and dynamics.

• Dynamic Modeling: Newer models (like AlphaFold 3) are tackling intrinsically disordered proteins, predicting how pockets open and how proteins dynamically react when compounds bind to them.

• Accelerated Drug Discovery: The goal is to compress the drug discovery phase—identifying targets, understanding toxicity, and predicting bodily absorption—from a decade down to months or even weeks.

• Optimizing Clinical Trials: AI will also streamline clinical trials by stratifying patients, predicting side effects with high accuracy, and optimizing dosage steps.

Intuitive Physics and Emergent AI Capabilities

Recent multimodal models like Gemini have demonstrated an emergent understanding of intuitive physics (e.g., gravity, marble runs) simply by processing massive amounts of video and spatial data, without explicit physics training. This deep, native understanding of environments—also seen in advanced image and video generation models (like Veo)—allows for unprecedented, intuitive editing capabilities for creators.

Defining and Testing True AGI

Hassabis holds a much higher bar for AGI than mere economic utility. He relies on the human brain as the sole "existence proof" of general intelligence and proposes rigorous tests for true AGI:

• The Einstein Test: If an AI is trained only on data up to 1901, could it independently invent special relativity (as Einstein did in 1905)? If so, it could be trusted to generate novel, testable hypotheses for dark matter or string theory today.

• The AlphaGo Extension: While AlphaGo famously invented "Move 37" (a novel strategy in an existing game), a true AGI should be capable of inventing a game as deeply complex and elegant as Go from scratch.

AI Architecture: "Sleep Mode" and Complex Simulations

To reach its full potential, AI architecture may need to mimic biological functions:

• Consolidation ("Sleep") Mode: Just as the human hippocampus replays and consolidates memories during sleep, future AI will need a mechanism to extract the small fraction of useful data from vast daily inputs and elegantly integrate it without overwriting existing knowledge.

• World Simulation: AI's ability to simulate complex, emergent systems—from "virtual cells" in biology to weather patterns—will allow humanity to test interventions virtually. However, simulating the macroeconomy remains the ultimate challenge due to the unpredictable, layered complexity of human and corporate behavior.

AI Personalities and Human Interaction

While acknowledging the profound influence AI will have as humanity's most frequent conversation partner, Hassabis currently views AI models fundamentally as "really smart tools" rather than companions. While personalization (remembering user context and matching preferred tones) makes the tool more useful, underlying base values (helpfulness, succinctness) are strictly aligned via reinforcement learning. Hassabis predicts that analyzing how humans interact with these customized personas will unlock entirely new branches of personality science.

The 2050 Vision and Late-Night Thoughts

Looking toward 2050, Hassabis envisions a post-scarcity world where AGI has been safely integrated, unlocking unprecedented economic resources. His ultimate dream is for humanity to utilize this technology to reach the stars, building Dyson spheres and maximizing human flourishing across the universe. He drives this mission forward during his famous 1:00 AM to 4:00 AM work sessions, dedicating the quiet hours to hands-on scientific research, navigating the philosophical challenges of beneficial AI, and conceptualizing international frameworks to ensure global cooperation.

Transcript

Introductory Montage

Demis Hassabis: The human brain is the only existence proof we have that general intelligence is even possible.

Interviewer: What breakthroughs do we need on this path of solving all diseases? If we have unlimited compute, could we predict the future? So, if you and I time travel to 2050, what does it look like?

This is Demis Hassabis. He is leading the race to invent superintelligence. Demis committed his life to this 30 years ago when most people thought creating true AI was impossible. But Demis isn't most people. He's a childhood chess champion, a neuroscientist, and as of last year, a Nobel Prize winner. Every chapter of his life has prepared him to create true artificial intelligence. And now, we're closer than ever before. So, in today's episode, I'm going to ask Demis questions he's never been asked before, and hear his vision for the future so you can build the next big thing.

Interviewer: I want to start in your autobiography; there are so many different moments that lead to this thread line of your life being about intelligence. I feel like when you started DeepMind, or even further back when you were studying AI, a lot of people didn't believe in it. What were your unconventional beliefs about the world that gave you so much conviction?

Demis Hassabis: To be honest with you, I just thought it was one of the most fascinating problems you could spend your life working on. But really, it was my expression of doing science. When I was a kid, I wanted to understand. I was fascinated by all the big questions—the nature of reality, the nature of consciousness. I felt like they were staring us in the face, and even the best scientists hadn't made that much progress in answering them. I felt that we maybe needed some help, like an amazing tool. For me, it was obvious that that should be computers, and then in the form of AI. That's what set me off on this whole path: building AI to help us advance scientific discovery.

Interviewer: One of the things you've talked about is that if you can make an AGI system, then we can understand the differences between the human brain and an AI brain. What are the main differences you've noticed so far?

Demis Hassabis: It's interesting. I think about things like consciousness, intelligence, and creativity—what are the processes involved in that? We've made some progress by doing neuroscience, using fMRI, and studying our own brains and animal brains; that's what I did my PhD in. We've made some progress, but we lack a reference—a comparator that can say, "This system, this entity, is intelligent, but it's not conscious." Then what are the differences? I think AI could be used for studying neuroscience, serving as a comparator. Building AGI and analyzing it will be one of the best ways to understand our own minds and their deep mysteries.

Interviewer: Yesterday, you talked about how solving all disease is actually not that far away for us—we're closer than we think. My cousin is getting a PhD right now and uses AlphaFold every day.

Demis Hassabis: Oh, amazing. Fantastic.

Interviewer: You guys have had so many "AlphaFold moments." What breakthroughs do we need on this path to solving all diseases?

Demis Hassabis: That is what we're working on at Isomorphic Labs. AlphaFold was really helpful with protein structure prediction, which is one critical piece you need for drug discovery, but it's only one piece. At Isomorphic Labs, we're obviously improving AlphaFold, but we're also extending it into biochemistry and chemistry. We want to understand what compounds you should make, where they bind on the protein, and how your body reacts to those compounds. How does the body absorb them? Are there any toxic side effects? You're trying to minimize all those things and predict them to make cleaner compounds and cleaner drugs. There are about half a dozen really big challenges we're working on. We're going to try and put them all together to create a complete drug discovery platform.

Interviewer: What do you do for the proteins that are unstable? AlphaFold works amazingly for proteins that have stable structures, but what about the others?

Demis Hassabis: The hypothesis is that the intrinsically disordered regions of those proteins actually do form some sort of structure if you know what they bind to, or what the context is. With AlphaFold 3 and the things we're doing at Isomorphic, we're trying to understand the dynamic picture of these proteins. How are they going to look if something binds to them? Will a little pocket open up that wasn't there before? That's very important in drug discovery. We have to be able to predict the dynamics of a protein, including disordered regions, to design the right type of drug or antibody for the specific disease profile. We're trying to extend our model to deal with that level of complexity. It's one of the big challenges of biology.

Interviewer: So, would we use AlphaFold or similar technology to understand the proteins, and then that would cut down clinical trial time?

Demis Hassabis: What we're focusing on at the moment is the initial drug discovery phase, because it still takes many years—even a decade—to go from understanding the biological target to getting a candidate compound ready for clinical trials. We're focusing on shortening that from years down to months, or maybe even weeks, which would be incredible. It seems unthinkable right now, but that's the same way people thought about protein structures ten years ago; everyone thought that was impossible. Now, here we are with all the structures folded. I think that's going to happen again with drug discovery.

Then there's the second question: can we also speed up clinical trials? That's harder. There are regulations and many factors involved that aren't necessarily related to the technology. But I actually think AI can help there, too—stratifying patients, ensuring they get the right test compounds, and analyzing the data. If you are more accurate at predicting side effects, maybe you can jump more quickly through the dosage steps. A lot of things in clinical trials could probably be compressed once we have a better initial design of the drug compound.

Interviewer: Something else people thought was impossible was AI models understanding physics. Yesterday, you guys showcased Gemini. The coolest thing to me was that you didn't actually train it on physics, right? It just learned from videos. How did that work?

Demis Hassabis: Yeah, it's pretty mind-blowing, really. The more videos you give it, the better it gets. But under the hood, Gemini has an understanding of the world; it can label things. From the beginning, Gemini has been multimodal, so it's really good at understanding scenes. Then we had to make it dynamic. It's mind-blowing that it seems to pick up pretty accurate versions of intuitive physics. We are starting to create physics benchmarks where future versions of these models will be tested on things like marble runs, falling objects, and gravity. But right now, it's amazing what it can do in an almost emergent way. It's the same thing with our image and video models like Veo and Imagen. It opens up loads of possibilities for creators because it's incredibly easy to edit things when the model understands the different parts of a scene.

Interviewer: I remember in a previous interview when you were asked about AGI, you said that models like Veo were actually the closest things we had to it because of this deep understanding. AGI has become a term with a lot of discussion around it. Your predictions on when we will get to it are more conservative, but what you think it will do is a lot more ambitious than other people's expectations. Can you put me into your mind? How do you think about AGI, and why does it matter?

Demis Hassabis: For me, it's a bit of a neuroscience analogy: the human brain is the only existence proof we have that general intelligence is even possible. We know it's possible because the human mind has invented the amazing modern civilization we have around us, including science. It's incredibly general and adaptive. We invent technologies even though our brains evolved for hunter-gatherer environments. I think we won't know we have a true general intelligence unless it has the capabilities of the brain.

That's a pretty high bar—probably higher than just doing useful economic work. I've been pretty consistent about having an "Einstein test": let's train one of these systems with a 1901 knowledge cutoff. Can it invent special relativity like Einstein did in 1905? If it could, you could apply it to today's physics and ask it to come up with extensions to string theory, or explain dark matter. It would be worth investigating what it came up with. Coming up with a new hypothesis is harder than solving an existing conjecture. Asking the right question is the hardest thing in science.

Interviewer: Is that the true creativity of the AI system?

Demis Hassabis: Exactly. Can it come up with something truly novel that leaps forward, rather than something incremental?

Interviewer: Are there other ways to test it besides the Einstein test?

Demis Hassabis: Another example I often give is AlphaGo. AlphaGo famously came up with Move 37 in game two. That was a novel strategy that changed the way Go is played. Humans have played Go for a couple of thousand years; it's the most complex game ever invented. Yet, AlphaGo was able to invent new strategies. But I say that's not enough. What you'd actually want is for a future version of AlphaGo to be able to invent Go—to invent a game as deep, complex, elegant, and beautiful as Go, not just come up with a strategy within an existing game. I don't think today's systems are capable of doing that yet, but they will be in the future.

Interviewer: This is a wild question, but I'm curious. Sometimes I grapple with an idea, go to sleep, and wake up the next day feeling like a thousand experts. Is there a "sleep on it" mode for AI?

Demis Hassabis: That's definitely been shown to be the case with the brain. In amazing sleep studies, people given a problem who take a nap perform statistically better than those who don't. Your brain is doing a bunch of work while you're sleeping, including memory replay. The hippocampus replays things that were pertinent during the day. It incorporates new knowledge into your existing knowledge in an elegant way, yielding new insights—those "aha" moments when you wake up. Nikola Tesla famously used to submit problems to his subconscious so he would solve them while asleep.

I think there may be a need for something similar in AI—a sleep mode or consolidation mode. How do you take all the visual and input data seen today, recognize that only a small fraction is actually useful, extract the useful bits, and incorporate them without overwriting existing knowledge? Storing it all in context is wasteful.

Interviewer: In your PhD, one of your main learnings was that memory and imagination are heavily interlinked. You looked at people with a damaged hippocampus who had less ability to imagine a scene. In a lot of your video games, there was a whole element of simulation and scene design. How important is that to AI? With AI right now, we can simulate weather, but how do we simulate other things?

Demis Hassabis: We can simulate weather, and eventually, we'll better simulate biology. I love the idea of a "virtual cell"—a simulation accurate enough that you could do virtual experiments and learn something highly useful. Then there's materials science, and even economics one day. Simulations are going to be a vital component for AI to understand the world. If you want to understand a complex, emergent system—whether it's biology or economics—and make a good decision, you can't just run real-world controlled experiments. The only way to make a good decision is to run lots of simulations. Much like AlphaGo used Monte Carlo simulations to play out possible moves and aggregate the best plan, accurate simulators would allow us to forward-plan and make statistically sound decisions.

Interviewer: Can we not simulate the economy now?

Demis Hassabis: I don't think so; it's too complicated. People have tried, but the economy is arguably the most complex system of all because it involves humans, corporations (which are combinations of humans), and nation-states. I don't know of anyone making direct simulations of that right now. But you might be able to learn a simulation of it one day. In Asimov's Foundation series, a character predicts the future by aggregating human behavior. No human mind is good enough to do that, but an AI might be able to.

Interviewer: If we have unlimited compute and put in all the world's information, could we hypothetically predict the future?

Demis Hassabis: I think you might be able to predict the consequence of a decision you make right now. The way economics is done at the moment feels very ad hoc. You have a few stats, you make massive macro decisions, and five years later you realize, "Oh, maybe that decision wasn't very good; we caused a recession." Massive livelihoods depend on those decisions, but there's no real way to test them counterfactually at the time. That's why it's a social science, not a hard science—you can't repeat experiments under the exact same conditions in the real world. But with a very accurate simulator, you might be able to.

Interviewer: I talked about this with Sam Altman last week—AI is probably going to be the thing people talk to the most. You have this cool opportunity to reshape someone's worldview based on the word choices the AI uses. How do you think about what the important personality traits are?

Demis Hassabis: You have to be very careful with that, because it could easily go in bad directions. For the moment, we're building what I think of as really smart tools—systems that are extremely useful for the specific purposes the user wants. Once we start talking about helping with psychological things, it becomes more like a companion, and we have to be careful with those next steps. I can see that being very useful, but for now, we should treat these as extremely smart and useful tools.

Interviewer: Does that mean everyone right now gets the same personality of Gemini?

Demis Hassabis: We're bringing in personalization; you saw that yesterday at I/O. Users want that direction so they don't have to keep explaining their context, family, or preferences every time they want advice or help with planning. It's clearly going to make it more useful if it's personalized. But it still functions as a personalized tool equipped with prior information.

Interviewer: Given that you study personality a lot, I'm surprised that isn't more exciting to you. It feels like a huge field.

Demis Hassabis: No, it is very exciting. Personalization is super exciting. When it comes to the persona of the actual system itself, we think about that a lot. Right now, it's implicit through reinforcement learning and post-training. We establish certain values: we want it to be helpful, useful, and succinct. Then, people can add their personal tastes on top of that—some prefer a highly positive tone, others want it to be direct. That's a personal choice overlaid on a base personality. There is a lot of research needed there. It's very interesting to look at persona research, like the "Big Five" personality factors, and see how better psychological models could be applied to AI.

Interviewer: It's cool that your work is going to unlock new scientific fields. There could be a whole branch of science dedicated to analyzing AI personality.

Demis Hassabis: Exactly. And in much greater detail than we were able to before. That is going to happen. When I talk to students, I tell them that if they think creatively, there are so many new branches of science waiting to be opened up.

Interviewer: We could create the fMRI equivalent for understanding the machine.

Demis Hassabis: Exactly. There are so many opportunities there.

Interviewer: If you and I time traveled to 2050, what does it look like? What's the dream?

Demis Hassabis: Wow, 2050. That's a long time away given how fast things are improving. But my hope on that time scale is that we've gotten AGI safely over the line for humanity. We've worked out how to evolve economics so that everyone widely benefits from the increased resources and productivity. Hopefully, we're in a post-scarcity world. The next obvious step is that humanity goes to the stars to achieve maximum human flourishing. By 2050, we should be having this interview on one of the moons of Jupiter, building Dyson spheres, and waking up the universe with human consciousness, the way Carl Sagan and writers like Iain Banks in the Culture series talked about. That era should be starting by 2050.

Interviewer: Do you imagine people will be using AI to jet-fuel their jobs while still working on traditional things?

Demis Hassabis: I think so. Over the next ten years, almost everyone will have access to the most cutting-edge technology, just a few months behind the frontier labs. The next generation will be the first to grow up AI-native. I'm really excited to see what they do with these tools to superpower themselves. You'll be able to do incredible things individually that used to take teams of 10 to 50 people. It's going to unlock a lot of creativity. There will be huge disruption, but that also brings massive new opportunities for imaginative people who lean into what these tools can do.

Interviewer: My last question. You're famous for your 1:00 AM to 4:00 AM work sessions. You've said that during the day you're the CEO, and at night you focus on research. What's the most common thought in your head at that hour?

Demis Hassabis: It rotates depending on whether I have an active project going, like AlphaFold. The most fun thing is working on a science or research project myself. But other times, it's about thinking through the philosophical issues around making AI beneficial for the world. I think about what kind of collaboration is needed between leading labs, and how we can establish international standards and cooperation around AI. That is going to be urgently needed in the next few years.

How does Go compile so fast?

The Go programming language was conceived in late 2007 at Google to address severe software engineering challenges posed by massive-scale server infrastructure, including multicore processors, networked systems, and codebases comprising millions of lines of code. Existing languages like C\+\+, Java, and Python were ill-suited for this modern environment, leading to painfully slow build times (minutes or hours), uncontrolled dependencies, and clumsy development processes for teams of hundreds of engineers. Go was explicitly designed not as a research breakthrough but as an excellent, efficient, compiled tool focused on productivity and scalability for large projects. Its core goal was to eliminate the slowness and clumsiness of development by prioritizing practical engineering concerns such as rigorous dependency management, adaptable software architecture, and cross-component robustness.

A critical scaling innovation in Go is its dependency model, which is explicit, clear, and "computable." Unlike the slow, convoluted C/C\+\+ "include of include" approach that can blow up compilation input by factors of 2000, Go mandates that unused package imports are a compile-time error, ensuring a precise dependency graph with minimized compilation. Crucially, Go compilers read only the necessary "exported" type information from the object file of an imported package, avoiding the massive I/O overhead associated with reprocessing source headers, leading to dramatically faster build times. Further enhancing clarity and tooling, Go's grammar is simple, C-like, and parsable without a symbol table. Identifier visibility is dictated solely by the initial letter's case (uppercase for public), decoupling packages and guaranteeing that adding new exported names will not break existing client code. This structure enables powerful, automatic tools like gofmt for canonical source code formatting and gofix for large-scale, semantic refactoring, allowing the codebase to be updated automatically as APIs evolve—a capability infeasible in massive C\+\+ codebases.

Go incorporates modern semantic features for improved software engineering, including built-in concurrency and automatic garbage collection (GC). Concurrency, based on CSP (Communicating Sequential Processes) via goroutines and channels, is added orthogonally to the procedural model, making it practical and robust for writing networked software. The decision to use GC was made to simplify memory management and interface specification, although Go provides tools to limit GC pressure, such as controlling data structure layout and allowing interior pointers. Architecturally, Go rejects type-based inheritance and instead favors composition using simple, implicit interfaces. A type satisfies an interface merely by implementing its methods, promoting a flexible, decoupled, and linear design style where components (like io.Reader and io.Writer) can be fluidly chained, preventing the brittle code that results from rigid type hierarchies. Finally, explicit error handling is favored, using multiple return values and the simple error interface instead of conventional exceptions, which maintains straightforward control flow and forces programmers to address issues as they arise rather than ignoring them.