Most people look at artificial intelligence and stop at chatbots. Words, sentences, clever answers. That’s where the fascination ends. But Eric Schmidt, who once ran Google, sees something else moving underneath, something harder to notice and maybe more important.
He recently spoke at a TED event, and what he pointed to wasn’t flashy. Not language. Not summaries. Planning. Strategy. That’s the shift. While the crowd is watching the show, he says, the real tech is learning how to think a few steps ahead—and do it alone.
Image: Ted / YT
Systems built on reinforcement learning, a method that lets them try, fail, adjust—are getting sharper. This isn’t about chat anymore. It’s about machines figuring out how to act without being told what comes next.
Right now, he says, everyone still thinks in terms of text. But AI is already going further. First it was language. Then came sequence—useful in biology, for instance. Now it’s about forming plans, solving problems in layers. And soon? Machines running whole business operations on their own. Quietly.
He’s using some of them, personally. After investing in the space industry, he wanted to understand it better. Not casually—deeply. Instead of reading textbooks, he asked AI systems to study for him. One spent about 15 minutes pulling together a dense, research-heavy writeup. Schmidt called it “extraordinary.” Not polished, but insightful. Not written for style, but for depth.
To fix that, LeCun proposed a new kind of model—H-JEPA. Its goal? Let machines figure out steps, not just sentences. The idea is that, to be smart, an AI has to work with goals. It needs to try, adjust, move forward. That’s different from just guessing the next word.
And Schmidt agrees. Without that kind of reasoning, it’s all just smoke.
Trade restrictions have already changed how China approaches AI. With limited access to advanced chips, they’ve been pushed to rely on algorithmic workarounds—code over hardware. In a way, that’s making them faster.
Schmidt claimed that discussions about AI are happening inside military circles—real ones. Strategic ones. He even mentioned that in some rooms, people have suggested preemptive moves. That’s not science fiction. He thinks it’s five years away, maybe less.
But he warned: this isn’t something to watch passively. “Ride the wave,” he said. Not once. Every day. The idea is, if you’re not using it, someone else is—and they’ll move faster than you.
In his mind, this isn’t like electricity or the internet. It’s bigger. AI, especially general intelligence, will shape the next century—or millennium. And it’s already starting to.
Read next: Future ChatGPT Could Store and Analyze Your Entire Digital Life
He recently spoke at a TED event, and what he pointed to wasn’t flashy. Not language. Not summaries. Planning. Strategy. That’s the shift. While the crowd is watching the show, he says, the real tech is learning how to think a few steps ahead—and do it alone.
Systems built on reinforcement learning, a method that lets them try, fail, adjust—are getting sharper. This isn’t about chat anymore. It’s about machines figuring out how to act without being told what comes next.
Right now, he says, everyone still thinks in terms of text. But AI is already going further. First it was language. Then came sequence—useful in biology, for instance. Now it’s about forming plans, solving problems in layers. And soon? Machines running whole business operations on their own. Quietly.
It’s Already in Use—Just Not Where Most People Are Looking
While people play around with bots that write poems or emails, deeper tools are being built. Schmidt mentioned a few: o3, Deepseek R1. They don’t just give answers. They try things out. Go back. Rethink. They loop.He’s using some of them, personally. After investing in the space industry, he wanted to understand it better. Not casually—deeply. Instead of reading textbooks, he asked AI systems to study for him. One spent about 15 minutes pulling together a dense, research-heavy writeup. Schmidt called it “extraordinary.” Not polished, but insightful. Not written for style, but for depth.
Experts Say Planning Is Where Intelligence Starts to Look Real
Schmidt isn’t alone here. Yann LeCun from Meta has said similar things—today’s large language models don’t really think, he argues. They don’t understand space, memory, consequences.To fix that, LeCun proposed a new kind of model—H-JEPA. Its goal? Let machines figure out steps, not just sentences. The idea is that, to be smart, an AI has to work with goals. It needs to try, adjust, move forward. That’s different from just guessing the next word.
And Schmidt agrees. Without that kind of reasoning, it’s all just smoke.
The Stakes Go Beyond Tech—It’s Geopolitical Now
He also framed the issue in terms of global power. Countries aren’t watching from the sidelines. The U.S. and China, he warned, are in a kind of AI arms race.Trade restrictions have already changed how China approaches AI. With limited access to advanced chips, they’ve been pushed to rely on algorithmic workarounds—code over hardware. In a way, that’s making them faster.
Schmidt claimed that discussions about AI are happening inside military circles—real ones. Strategic ones. He even mentioned that in some rooms, people have suggested preemptive moves. That’s not science fiction. He thinks it’s five years away, maybe less.
Still, He Sees Room for Hope—If People Keep Up
Despite the tension, Schmidt isn’t all doom. He still sees AI as something that could change lives—in education, medicine, material science. He talked about AI tutors that adapt to the person, tools that help researchers explore what’s still unknown.But he warned: this isn’t something to watch passively. “Ride the wave,” he said. Not once. Every day. The idea is, if you’re not using it, someone else is—and they’ll move faster than you.
In his mind, this isn’t like electricity or the internet. It’s bigger. AI, especially general intelligence, will shape the next century—or millennium. And it’s already starting to.
Read next: Future ChatGPT Could Store and Analyze Your Entire Digital Life
