The AI revolution is underhyped | Eric Schmidt

**Elise Hu** (0:07)
You're listening to TED Talks Daily, where we bring you new ideas and conversations to spark your curiosity every day. I'm your host, Elise Hu. AI and the future of humanity were huge topics at this year's TED Conference. Central to all of this was a pretty existential question. In today's world, what is a human actually for? To add to this discussion, Google's former CEO and Chairman Eric Schmidt joined creative technologist Bilawal Sidhu for a conversation about AI and our collective future. They discuss what, if any, are the limits of AI, ethical questions about its rising use across various sectors, and why the AI revolution, as Eric puts it, is underhyped.

**Bilawal Sidhu** (0:55)
Eric Schmidt, thank you for joining us.

**Eric Schmidt** (0:56)
Thank you.

**Bilawal Sidhu** (0:57)
You said the arrival of non-human intelligence is a very big deal. What did you see that the rest of us might have missed?

**Eric Schmidt** (1:04)
In 2016, we didn't understand what was now going to happen, but we understood that these algorithms were new and powerful. There was a new move invented by AI in a game that had been around for 2,500 years that no one had ever seen. Technically, the way this occurred was that the system of AlphaGo was essentially organized to always maintain a greater than 50 percent chance of winning. And so it calculated correctly this move, which was this great mystery among all of the Go players, who are obviously insanely brilliant mathematical and intuitive players. The question that Henry, Craig Mundy and I started to discuss is what does this mean? How is it that our computers could come up with something that humans had never thought about? I mean, this is a game played by billions of people. And that began the process that led to two books. And I think frankly is the point at which the revolution really started.

**Bilawal Sidhu** (2:08)
If you fast forward to today, it seems that all anyone can talk about is AI, especially here at TED. But you've taken a contrarian stance. You actually think AI is underhyped. Why is that?

**Eric Schmidt** (2:22)
And I'll tell you why. Most of you think of AI as, I'll just use the general term as ChatGPT. For most of you, ChatGP was the moment where you said, oh my god, this thing writes. And it makes mistakes, but it's so brilliantly verbal. Right, that was certainly my reaction. Most people that I knew did that.

**Bilawal Sidhu** (2:39)
It was visceral, yeah.

**Eric Schmidt** (2:41)
This was two years ago.
Since then, the gains in what is called reinforcement learning, which is what Alphago helped invent and so forth, allow us to do planning. And a good example is look at OpenAI 3 or DeepSeq R1, and you can see how it goes forward and back, forward and back, forward and back. It's extraordinary. In my case, I bought a rocket company because it was like interesting. And I know as one does, and it's an area that I'm not an expert in, and I want to be an expert, so I'm using deep research. And these systems are spending 15 minutes writing these deep papers. It's true for most of them. Do you have any idea how much computation 15 minutes of these supercomputers is? It's extraordinary. So you're seeing the arrival, the shift from language to language, then you have language to sequence, which is how biology is done. Now you're doing, essentially, planning and strategy. The eventual state of this is the computers running all business processes, right? So you have an agent to do this, an agent to do this, an agent to do this, an agent to do this, and you concatenate them together and they speak language among each other. They typically speak English language.

**Bilawal Sidhu** (3:59)
I mean, speaking of just the sheer compute requirements of these systems, let's talk about scale briefly. You know, I kind of think of these AI systems as hungry, hungry hippos. They seemingly soak up all the data and compute that we throw at them. They've already digested all the tokens on the public Internet, and it seems we can't build data centers fast enough. What do you think the real limits are, and how do we get ahead of them before they start throttling AI progress?

**Eric Schmidt** (4:27)
So there's a real limit in energy. To give you an example, there's one calculation, and I testified on this this week in Congress, that we need another 90 gigawatts of power in America. My answer, by the way, is, think Canada, right? Nice people, full of hydroelectric power. But that's apparently not the political mood right now. Sorry. So 90 gigawatts is 90 nuclear power plants in America. Not happening. We're building zero.