[0:00:00] If the new world is going to be run by artificial intelligence, what does it mean for humanity? Can people stay at the top of the food chain or will we be overtaken? I think that this is going to be an intelligence explosion unlike any trajectory of technology we've seen in history. This year has seen a seismic breakthrough in how AI uses language.
[0:00:28] Computers are now harvesting trillions of words from every imaginable source and using that to talk to us like they're other humans and triggering a wave of warnings about the existential threat to come. I'm a humanist. I think our goal should be to keep humans at the top of the food chain. That means we have to prevent proliferation and ensure that these tools remain completely accountable. What this man says matters.
[0:00:58] He's played a key role in the development of AI from its first critical steps to its latest incarnation of intelligent chatbots. Why can't we breathe underwater? Well, there are a couple of reasons. Three years ago, these models were barely able to string together a single coherent sentence. You now have a tool completely at your disposal 24-7 to teach you.
[0:01:25] And guide you, give you feedback, help you remember, brainstorm, be more creative. And I think that is going to be one of the most meritocratic turbo boosts to productivity in the history of our species. But there is also a darker side. Let loose, will these programs always align with us? Or could they be more like James Cameron's dystopian vision? Salomon says AI doesn't have to become sentient.
[0:01:54] To cause havoc for the modern world. There's been a hell of a lot of focus on the Terminator robots and I think there's a lot of practical near-term concerns that should preoccupy us. I think that's the risk is that people will use this to spread misinformation, to deliberately pollute the information ecosystem, to sow doubt, to sow fear.
[0:02:17] And, you know, potentially campaign against democracy. We've certainly seen, you know, the Russians and others to try and participate in that kind of thing. So I expect more of that. Hey, Pi, I'm thinking about taking a trip to London. Pi, created by Mustafa Suleiman, along with its rivals like ChatGBT, can answer technical, planning and even emotional questions.
[0:02:40] Every time you ask it something, the technology is learning, both from you and the data it collects. It looks like the Regent Park boat ride is open daily from 10.30 a.m. to 6 p.m. And its interactions are always adapting to fit you. Should AI have a moral compass? These models are tools. So people who use those tools will be able to insert their values into those models.
[0:03:08] And that's where it's likely to cause chaos and conflict. Denying that reality is putting your head in the sand. So does AI need to be contained with tougher regulation? Containment requires coordination and it requires restrictions on how we train the very largest models and how they're deployed in practice. Are you asking to be contained?
[0:03:30] Yeah, I think for sure. When you're building a technology that is as fundamental as intelligence itself, then we should be really sensitive and conscious about how those tools proliferate. Both the power and the vast sums of money that are to be made in AI have created an arms race for companies and countries that's hard to control. It is historic. The EU has a track record of clashing with American tech giants like Facebook and Google.
[0:04:00] But as new AI rules are signed into law, will this new generation of titans accept tougher regulation? I certainly will accept it. And I know that many of my peers who I've come up with at DeepMind and OpenAI are also very inviting and very welcoming of an experimental, proactive, interventionist regulatory framework. This is what we need. And we need to be more forgiving and more collaborative so that we don't shut things down and be damning of those who make mistakes.
[0:04:27] Just weeks ago, Sulaiman joined with other AI leaders to sign a stark 22-word statement reading, mitigating the risk of extinction from AI should be a global priority. If it's so risky, why don't you stop? I think that the point of raising concerns is that we can see a moment at some point in the future, probably over a decade or two decades time horizon.
[0:04:52] When slowing down is likely going to be the safe and ethical thing to do. 10 years is not a long time. Well, the decisions that we have to make are how we prevent bad actors getting access to technologies which can destabilize our world. Just as we don't allow, you know, anyone to experiment with nuclear materials, right? That is a licensed and restricted product. This is going to be exactly the same kind of environment.
[0:05:22] But arguably easier to get your hands on AI coding than on nuclear weapons. I think that's exactly why we're all raising the concern, because nuclear material, uranium-235, is a physical substance which is actually very expensive, very difficult to handle, and very concentrated. On the flip side, these tools will be open sourced and very widely spread around the world. But it's not just humanity that's under threat.
[0:05:54] I certainly think Google's, you know, been caught off guard in the last 12 months. I mean, Google's certainly been behind. And, you know, it's an incredible company, so I'm sure it will do just fine and catch up. But I certainly think this is a moment when the incumbents...
[0:06:13] Are threatened because many, many other companies now have the tools and indeed many academic labs and open source projects to try to reimagine the way that you interact with computers, which is why it's such a creative and exciting time. We really will be speaking to computers in the next few years. If the future of computing is conversational, does that mean that it's all over for search engines? It's all over for Google?
[0:06:38] I think that the future interface is for sure going to be conversational. The way that you're going to get things done online and the way that you're going to learn things, buy things, book things, is that you're just going to ask your personal AI. Silicon Valley, where Suleiman is now based, is a long way from the modest North London neighborhood he grew up in. He recognizes he's an optimist about the technology, but he says he's not blind to the risks it poses.
[0:07:07] The question is how others will exploit or contain this new and growing intelligence.