 It's always embarrassing when people introduce me. I know half of it's made up. Right, we're going to talk about AI. We're going to get everybody on the same page in terms of what AI isn't. We will talk about how we can bring people and technology together to do amazing things. And we'll talk about the macro impact that these technologies are going to have on society. So a little bit about me. So I wear two hats. For the past 25 years, I've been researching AI. My undergraduate and my PhD postdoctoral in AI. I'm an entrepreneur in residence for UCL, which is one of the world's leading AI universities, where I help them spin out deep technology companies. I started a company 15 years ago during my PhD. There's been building solutions for big companies. And we were acquired by WPP two years ago when I'm the chief AI officer. So as me, it's boring. Before we get into definitions of AI, what I want to do is take you through this technology stack. This is how I think about building data-driven decision-making systems. I'm going to make this very interactive. I'm going to ask you lots of questions. There may be time at the end for some questions to be asked for me. But the first question I want to ask everybody is, what is data? Does anybody know the definition of data? It comes from the Latin word datum, which means given things. It's stuff. It's a fabric of our universe. Bits and bytes, ones and zeros. It's the sound that's going into your ears. It's the light that's going into your eyes. If I say to you 210280, this is data. If I say that it's a date of birth, or how much money I have in my account, then it becomes information. Information is data in context. So if I say to you that my laptop is sick, what does that mean? My laptop is sick. Are you cool? Or you're older? If you're cool, it means it's awesome. If you're old and boring like me, it means it's broken. So that sound that's going into your ears can be interpreted in two different ways. It's very important that you interpret it correctly. If I asked you to read this sentence here and create an image in your mind of what you think is happening, so read the sentence. What picture do you have in your head? Anybody brave enough wants to share with me what's going on in their head? Yes? John is reading a letter out loud to Mary. This can be interpreted in many ways. John could be reading the letter that was addressed to Mary, or sent to Mary, past tense. Doug could be reading the alphabet, ABCD to Mary. He could be reading the letter of the law to Mary. He could be reading a book called the letter to a lady called Mary. He could be reading a book called the letter to Mary. It could be an instruction. John, read the letter to Mary. This can be interpreted in many, many different ways. We will all go away with a picture in our mind of what we think is correct, and we're probably wrong. And my goal today is to convince you all that humans are rubbish. That's my goal today. Imagine if I give you a spreadsheet. And in the spreadsheet, it has three columns. It has a column, which is the date, which is the temperature on those dates, and a column, which is a number of ice creams that I sold on those dates. We've got dates, temperature, ice cream cells. What would you do with that spreadsheet? What would you do with that information? Maybe draw a graph, find some insights. Maybe draw a graph. So along the x-axis, we have increase in temperature. Along the y-axis, we have increase in ice cream cells. If we plotted that, we'll see some sort of trend going upwards. The hotter it gets, the more ice creams we sell. This is called descriptive analytics. What we're doing is we're organizing information to try to find patterns, to know things about the world. Plato had a very good definition for knowledge. He said, knowledge is justified through belief. Knowledge isn't a fact. Based on our statistical experience, we believe that we can do something to be true about the world. Now, we can do something really cool now, which is what everybody's excited about. We can do a thing called predictive analytics. And we can put a line through it. So I know you're all data scientists. Congratulations. We now have predictive power. And we can use more and more sophisticated mathematics equations to get a better fit of that line to that data. I would argue that all we're really trying to do in data science or machine learning is to fit some sort of line or plane or hyperplane to some data set to give us the best predictive power. That's really all we're trying to do. Knowing something is very different to understanding something. So if I asked you, why do we sell ice creams when it's hot outside? People say, well, it cools you down. Well, in the UK, when it's hot outside, everybody goes outside because it's a scarce resource. We only see the sun for two weeks of the year. Then we get hot. We can cool ourselves down in different ways. We can go back inside or we can take our clothes off. We can buy an ice cream. And why does the ice cream cool us down? Thermodynamics. We expect a computer to understand that narrative from this line is impossible. It doesn't understand about thermodynamics. It doesn't understand about human behavior. So it's really important that you bring the people that have domain knowledge, that people understand the world together with the people that are very good at extracting patterns. This does not happen enough in industry because your domain experts are going to say, Daniel, the model of the world here is wrong. It's existentially wrong. Because if tomorrow is a hot state ever, scorching hot, my factory is going to manufacture lots of ice creams. But the reality is that nobody will go outside because it's too hot. So the world looks more like this. So I would argue that what we're ultimately trying to do is understand the world perfectly so that we can make good decisions. It turns out that humans are rubbish at making decisions. I don't know if you've read the book by Daniel Kahneman thinking fast and slow. Daniel Kahneman is a Nobel Prize-winning economist. And he argues we have a fast brain and a slow brain. There are parallels in terms of how these brains work with how AIs work. Now, I'm going to test you all now. I'm going to ask you some math questions. I'm going to test your brains. And your job is to answer these math questions as fast as possible. If you don't answer them faster than your competitors, they're going to take your market. So speed is important. Don't be shy. We're going to start out nice and easy. So the first question is what's 2 times 2? Good. Fantastic. What's 14 times 2? OK. So the point here is that this one used your intuition. You used your fast brain. This one, you have to go through a process and algorithm to come up with the right answer. I'm going to ask you a few more math questions. Again, we'll start out nice and easy. We'll get more and more complicated. So the next question is the combined price of a bat and a ball is $1.10. The bat is $1 more than the ball. How much is the ball? If you said 5 cents, you knew the answer already. If you said 10 cents, you're not broken. That was your intuition getting the wrong answer. The bat is $1.05. The ball is 5 cents. The combined price is $1.10. The bat is $1 more than the ball. I can give you lots of these examples and we get them wrong. Because we're using our intuition to make decisions and we shouldn't be. Imagine these are five staff members, five employees. Ignore those rules on the right that are important. How many ways can I allocate five people to five jobs? How many ways can I allocate five people to five jobs? Remember that exclamation mark will enter school factorial five times four times three times three. So 120 possible ways to allocate five people to five jobs. Let's make the problem more complicated. I've got 15 people. How many ways can I allocate 15 people to 15 jobs? Don't say 15 factorial. It's cheating. Fast brains, anybody want to shout out a number? So a trillion possible ways. To expect a human to solve this problem, we're wasting our time. Anything more than seven, don't use a human for. That's a good rule to take away with you. And actually, industry don't have problems of this size. They've got problems of this size. So here are 500 staff members. How many ways can I allocate them to the big number? It's a number that's got over 1,000 digits. And just to put this number into context, this is how many atoms are right in the universe. Once you reach about 60 things that you have to consider, 60 people in this case, there are more possible combinations than there are atoms in the universe. Humans can solve problems up to seven. You can hire a good computer scientist and can solve problems with a 30 or 40. Beyond that, to solve problems at this scale, you need to have deep, deep specialized expertise in algorithms. Let's just draw this problem home. Imagine you've got your ice cream van and your ice cream van needs to deliver ice creams around these 24 points. Humans are actually quite good at solving these spatial problems. So after a few minutes, we'll draw a nice path around these points. How long will it take a computer to get the shortest path? The one that's going to save us the most amount of petrol. Anybody? Milliseconds. There was somewhere between that and 20 billion years. If you put 24 times 23 times 30 into your calculator, you get this number of roots. That's how many roots are around those 24 points. If you had a computer that could check a million roots a second, million roots a second, it would take 20 billion years to go through every single possible one and say, this one that I looked at 10 billion years ago, this one's the shortest one. If I had another point to the map, it's now 25 times 20 billion years, 5,000 billion years. Another point to the map is 26 times 5,000 billion years. These are exponential problems. They get ugly very, very quickly. Human beings are solving these or you're using algorithms that were developed 20 years ago that are also solving them badly. When I build systems for organizations that usually have these three parts, everybody's getting excited about data. They're all building data lakes. I would argue that's not the right thing to do. That's a different conversation. We're all then putting Tableau and some sort of analytic layer on top of it, thinking we have AI. We're hiring my students, my master's students or data scientists to extract insights from data, hoping that those insights lead to better decisions. They typically don't. Human beings are bounded by our decision-making ability. Giving human beings better insights does not typically lead to better decisions. I would argue what you need to do, solve the problem at the top, the optimization problem, which if you're old enough used to be called operations research, discrete mathematics. It's a different field in computer science. I would solve that problem first, work backwards, figure out what insights you need to make better decisions and what data contains other insights. Anyway, if I build a system that I give data to and it makes a decision and tomorrow I give it the same data, it makes the same decision, I have automation. An automation is amazing because we can get computers to do things better than human beings. But does anybody know the definition of stupidity? What's the definition of stupidity? Just doing the same thing over and expecting a different answer. I would argue that by definition, automation is stupid. Not that it's not valuable, it's very valuable, but by definition it is not intelligent and therefore not AI. I know that everybody that currently switches this technology stack calls themselves an AI company, it's fine, we get more clients, we get more VC funding. But there are lots of definitions of AI. The first, sorry, the most popular one is actually the weakest, as far as I'm concerned, which is getting computers to do things that humans can do. So over the past decade, we've managed to get machines to corresponding natural language, to recognize objects and images, and when we get machines to behave like humans, because humans are the most intelligent thing we know in the universe, we assume that that's AI. I would argue that humans are not intelligent, that's a different conversation. There's a much better definition of AI that comes from the definition of intelligence. So instead of using human beings, as definition of intelligence says this, which is goal directed adaptive behavior. Goal directed in the sense you're trying to achieve an objective, you're trying to route your vehicles to maximize deliveries, or allocate your workforce to maximize demand, or spend your marketing money to maximize reach, whatever it's complex optimization problem. Behavior is how quickly I can answer that question now. We've just discovered that most of these problems are exponential in size. If you choose the wrong algorithm, it will take longer than the age of the universe. If you choose the right algorithm, it will take milliseconds. But the key word in this definition is adaptive. What you want to do is build systems, then make decisions, learn about whether those decisions are good or bad, adapt themselves so that next time they make their decisions. If I'm being totally honest with you all, we don't really see adaptive systems in production. Most systems in production are our automation, but the true paradigm of AI are systems that can safely adapt themselves in production. A little bit of a history lesson before I talk about a different way of thinking about AI, and some of the macro concerns. This is AI in the 60s and 70s. This is Socrates before he drinks hemlock and kills himself. Socrates is famous for inspiring the Socratic method. So if I say to you, Socrates is a man, and all men are mortal, I can infer that Socrates is mortal. So we write down lots of rules, lots of things that we know about the world, and then try to infer new knowledge. And dimly scale, dimly work. And then in the 80s and 90s, a new type of AI came along that's modeled on how brains work. This is the brain of a bumblebee. My PhD 20 years ago was trying to take the brain of a bumblebee and model it in a machine. Bumblebees have a million brain cells. Their brains can fit on the end of a needle. They can do amazing things. They navigate 3D worlds, and they recognize objects. They talk to each other. They don't handle windows very well, but ultimately they're very smart creatures. And the question was, can you model a million neurons in a machine? 20 years ago, you couldn't. Now we can model billions of neurons in a machine. We call these large language models or generative AI. These technologies are really good at knowing things about the world. They're really good at even telling you what they know about the world through text and imagery. They are not good at reasoning yet, and they are definitely not good at complex decision making. Use these brains to know things, use these technologies to then reason and make decisions, and then build systems that can adapt themselves in production. If I hold up pretty much everything we do in industry to this definition, nobody's really doing AI, which is not very useful, because actually the convergence of algorithms, compute data now enable us to do some really interesting things in our organizations. And I think there are six applications of interesting things. So the first category is task automation. So we can use very simple algorithms, if then else statements, macros to essentially free up mundane repetitive tasks that humans are doing. And whilst these technologies are simple, you can drive a huge amount of value in your organization. Now actually I'm doing a session tomorrow, talk much more about these applications to marketing tomorrow. So just with through them. Content generation, of course, everybody's excited now about content generation. The challenge is not generating general content. The challenge is how do you generate brand specific content? I'll talk more about that tomorrow. Human representation is where we can replace human beings that are the interface with things that look and behave exactly like a human being. And again, I'll talk about that more tomorrow with regards to large language models and how that changes the whole landscape with human representation. Insight extraction is what we've been excited about for the past decade. We've been using machine learning, data science to extract insights from data, hoping that leads to value. It typically doesn't. But what these technologies are really good at is not just predicting the world by explaining how the world exists. And we can use that explanation to actually make better decisions. Complex decision making we've already talked about. And then finally, human augmentation. So we can build essentially digital representations of you, your employees using large language models. And we can use that digital twin and ask you, if I put you on this project, will you work well? If I put you on this team, will you thrive? That's a real life example. And I love this framework, not just because we invented it in WPP, but it helps you navigate the complex world of implementing these technologies safely and securely and ethically. So this is a framework that we use to navigate implementing those. The types of questions that you need to ask yourself when implementing task automation are very different when implementing digital twins of employees where you can identify secret lovers and people who are going to leave the company before they know they're going to leave the company. All right, let's talk about the macro impact of the technology. So I don't know if you know who said this quote at the top, the nation that leads in AI will be the rule of the world. Does anybody know who said that? It's Vladimir Putin recently. So Vladimir Putin, so I think we need to acknowledge that these technologies are not just going to have the most profound impact on society, they're going to have the most profound impact on humanity. And I think we've all heard of the world singularity. You're all linked in AI philosophers now. A singularity comes from physics. It's a point in time that we can't see beyond. And it was adopted by the AI community to refer to the technological singularity, which is the point in time where we build a brain a million times smarter than us. I think there are at least six singularities that we need to concern ourselves with. Very quickly, the political singularity is where we no longer know what is true. Deep fakes, misinformation bots have not only challenged our political foundations and continue to challenge our political foundations, but they're now challenging the fabric of our reality. I know people that are being attacked by deep fakes or clones of their children, of their peers. I actually think that this industry will play an important role in ensuring that all content is authenticated and ensures providence. Environmental singularity is something we're all familiar with. We know that consumption is putting pressure on our planetary boundaries. I believe if we apply these technologies in the right way, we can halve the amount of energy that we require to run the world. And I can't go into the details about that during this session. Maybe I'll talk about it more tomorrow. The social singularity is not my expertise, but there are scientists that believe that there are people alive today that won't have to die. AI is advancing medicine. It's able to monitor ourselves and clean ourselves out. And a bit like a car, if we stay on top of damage, that car will never ever break down. The technological singularity is what we're concerned about at the moment, which is building a brain a million times smarter than us. This is the last invention that humanity will create. We have no idea whether it's gonna be the most glorious thing that happens to us or our biggest existential threat. My community thought it might not happen for another 30, 40 years. We now think it might happen in the next 10 or 20 years. If this thing comes along and sees us as a threat, it might remove us from the equation. So when it does, look busy, be nice to each other and hopefully it will go away. The legal singularity is when surveillance becomes ubiquitous. So this is the concern that AI now not only is able to know and profile consumers, people very well, but they can also manipulate their behaviors. And I believe, again, this industry will play a significant role in mitigating the risks of abusing these technologies. Finally, my favorite singularity, which is coined by a very good friend of mine called Callum Chase. This is the economic singularity. This is the concern about job losses. And I think over the next decade, we are gonna see a camera in explosion of innovations. Yes, jobs will be displaced, but new opportunities for work will appear. I think in the next 10 years, this is gonna be glorious, we're gonna free loads and loads of people up to be able to contribute to the world. There is a school of thought that actually we should be accelerating towards this singularity, that we should be removing friction from the creation and dissemination of food, healthcare, education, shelter, transport. If we can remove the friction, which usually means human labor, we can bring the cost of those goods down to zero. Imagine being born into a world where you don't have to work to pay for food, healthcare, education, it's all free. Giving you the ability to do whatever you want. Now, people say to me, Daniel, what would I do if I didn't have to work? I know lots of people who don't have to work, they're not sitting at home bored and depressed. They use their time and energy to contribute to humanity. And I bet if I asked you all, what would you do if you didn't have to work? You'll say, I'll travel, I'll indulge in my hobbies, I'll spend more time with my friends and family. And I bet if I keep pushing you, most people say the same thing, which is you want to do something that contributes positively to humanity. I believe that AI has the ability to create a world of abundance, freeing people up to actually contribute to humanity. So on that note, and we all know this, it's not good enough anymore for organizations just to have a strong, profitable business. You need to have a purpose. If you don't have a strong purpose, you're not going to attract talent, you're not going to attract customers. WPP's purpose is to free people up to enable them to create to make a better world. I believe a better world is a world where everybody is economically free to do whatever they want. And I actually believe it's enterprise that will make this world. I believe it's the collective purpose of all of you, of enterprise, that will make the next 10 years amazing. So on that note, I'm going to shut up and stop. Thank you. Thank you.