The fate of artificial intelligence, and perhaps the world, lies in the hands of a man born to a Syrian taxi driver and an English nurse. It’s not the plot of a dystopian movie. It’s the real life of Mustafa Suleyman, one of the world’s 100 most influential people in the field of AI.
At almost 40 years old, Suleyman can boast an extraordinary CV: in 2010 he founded together with Demis Hassabis one of the most advanced artificial intelligence research laboratories in the world, DeepMindwhich Google acquired in 2014. In 2022, however, Suleyman founded Inflection AIa company so influential in the field of AI that it deserved to be summoned to the White House.
Finally, last March Microsoft has chosen Suleyman to lead its artificial intelligence. The work and decisions of this AI pioneer, therefore, will profoundly influence how the company with the highest market value on the planet develops revolutionary technology in the coming years.
As leader of a tech superpower, Suleyman has enormous responsibilities. Not just from a business perspective. People in similar positions to him do not fall asleep easily. Sam Altmanthe CEO of OpenAI, said some time ago that he lost sleep over “the hypothetical idea that I had done something very bad the moment ChatGpt was launched.”
ChatGpt it is the most famous generative AI in the world. One hundred million users use it every month to generate texts and images similar to those that a human being would produce. In this technology, created by OpenAI, Microsoft invested 13 billion dollars.
In addition to increasing human productivity, ChatGpt has helped spread a growing concern: Machines will one day be so intelligent that they will wipe out humanity. Like a gigantic wave crashing into helpless swimmers.
Mustafa Suleyman talks about this tsunami in his book “The wave that will come”published in Italy by Garzanti.
“A new technological wave is about to crash around us. A wave that will give us the power to build the universal fundamentals of intelligence and life” writes Suleyman. But artificial intelligence, which promises to generate enormous wealth and fight diseases and the climate crisis more effectively, it is difficult to govern. At the moment, Suleyman says, it doesn’t seem possible to “control it, slow it down or stop it.”
“The thing I want most is for someone to prove me wrong and show me that containment is actually possible,” Suleyman writes in his book. Now that he works in the most powerful company in the world in the field of AI, Suleyman’s mission could be precisely to deny himself.
How do you think you can contribute to an AI containment strategy by playing a key role in Microsoft, a company that aims to commercialize this technology as much as possible?
“I believe you can’t get anything done unless you are in a position to truly shape and influence outcomes. In the book I underline precisely this: we need critical and responsible voices from within. And that’s what I can do at Microsoft: create AI-enabled products that have a containment philosophy at scale.”
Geoffrey Hinton, one of the pioneers of modern AI, left Google to be able to speak more freely about the risks associated with this technology. You wrote “The Wave to Come” before you were hired by Microsoft. How can we be sure that in the future, if necessary, the alarm will sound again?
“I am working to produce safe, ethical, responsible and trustworthy AI. This is what I came to Microsoft to do. And this is what I have always aimed for.”
Ever since ChatGpt was opened to the public, I’ve had a recurring thought: Developers see enormous dangers in AI because they have access to unfiltered or uncensored versions of the AI models they’ve created. It’s correct?
“Not at all. What developers see is in line with what all users can see. When people talk about long-term dangers, they do so by extrapolating their beliefs from what exists today. They are not based on the current capabilities of the AI models, but rather on where they think these models will be able to go in a given period of time. I believe everyone should listen to these voices with more humility. We don’t know exactly what will happen. And that’s a compelling reason to be cautious not just about the future itself, but also about what we say about it [dei pericoli legati all’IA]”.
Until a few months ago, the main concern for many was that AI could exterminate humanity. Today, however, the fear is spreading that AI may just be an economic bubble. They are right?
“Look around. Look at the world we have built. First and foremost, everything around us is affected in some way by technology. Right now I’m typing on a computer in a room built with power tools, lit by electric light, and so on. Every aspect of our lives is conditioned, at some level, by an infinite number of technologies. Now also consider how human intelligence has shaped and directed everything around you. Well artificial intelligence combines both of these forces. That’s why it’s so important. I’ve been working on this technology for 15 years and have seen interest fluctuate at times. But I have no doubt that this technological change is fundamental and, moreover, still underestimated.
In the book “The Wave to Come” you provide many examples of the capabilities of AI. This struck me: within three to five years, the book says, AI could potentially generate, in just a few months, a million-dollar profit from an initial investment of $100,000. Will we all get rich?
“That’s the idea! The goal of such AI would be to make everyone much more productive and capable, leading to a permanent increase in economic growth and personal wealth around the world. It’s a process we’ve seen before, starting with the industrial revolution. It has led to a great leap forward in well-being. If we do things right, I believe AI might be able to do something like this once again.”
The problem is that even if we get rich – if you are right about most of the dangers you talk about in the book – we won’t be able to truly enjoy that money. Among the many dangers you described, which one worries you most at the moment?
“I don’t like to isolate just one risk. Instead I would emphasize the idea of containment as a holistic approach to mitigate all possible harm. AI is a general purpose technology and that means its applications are broad, which is why it’s so exciting. But this also means that any response must be equally broad. So let’s think about the dangers broadly.”
Among the people most scared of “out of control” AI is Elon Musk. Would you ever work for (or with) him?
“It’s an interesting hypothetical question. I won’t say anything else.”
Media is at risk of being overtaken by generative AI. Do you have any advice to offer editors and journalists on how to survive?
“I would reverse the perspective. Media have an incredible opportunity to find new audiences and create new forms of reporting and content that are hugely relevant to large numbers of people, using AI. My advice is to experiment, learn, be bold and create tomorrow’s content, products and tools for readers.”
Big tech companies provide consumers with their most advanced AI models, but in most cases ask consumers for a cash subscription. Yet, these same companies use user input for free to train their AI. Isn’t user data enough to pay for the technology being offered to them?
“I’m not sure that’s true. Many major AI models are available for free, right now. You don’t need to pay to use powerful versions of software from companies like Google, Microsoft, Anthropic and OpenAI.”
[Suleyman ha omesso il fatto che le versioni più potenti delle IA delle aziende che ha nominato, sono tutte a pagamento. Le versioni free esistono ma sono molto limitate nelle prestazioni e nelle funzioni disponibili, nda]
In July 2023, President Biden convened representatives of the world‘s most influential AI companies. She was there too, and you were all men. Is the face of AI male?
“It is disappointing that there wasn’t better representation and balance in that fight and others like it.”
Mustafa Suleyman at the White House, where he met President Biden. Suleyman is standing fourth from the left.
You and Demis Hassabis, the current CEO of Google DeepMind, grew up together and now find yourselves at the top of the world, as rivals. Are you still in contact?
“We are on good terms. Demis invested in my latest company [Inflection AI, nda] and we’ll update when we’re both in town.”
Some of his colleagues who lead leading AI companies have said that this technology keeps them awake at night. Does she sleep well?
“I work very hard to make sure this technology is responsible and brings enormous benefit to humanity. Security, trust: these are fundamental things for me and are reflected in everything I do. It’s something non-negotiable for me, and I wouldn’t do anything if I had to give it up.”
v