AI won't need to take orders from humans soon: Ex-Google boss Eric Schmidt

Despite AI being in the news everyday, Eric Schmidt maintains most people still do not understand the scale at which artificial intelligence is growing

Artificial Intelligence is fast approaching a point where it may no longer need human input to evolve, former Google CEO Eric Schmidt warned. Schmidt also raised questions about the role of humans in a rapidly shifting technological landscape.

 

Speaking at an event last week hosted by the Special Competitive Studies Project -- a think tank he founded -- Schmidt said we are on the brink of a new phase where AI systems not only learn and improve independently, but also replace highly skilled professionals.

 

"The computers are now doing self-improvement… They’re learning how to plan, and they don’t have to listen to us anymore," Schmidt said. He described what he called "recursive self-improvement," a cycle in which AI systems generate new hypotheses, test them using robotic labs, and feed results back into the loop. All of this is done without human involvement.

 

 

“This is the future model of the fusion of AI and bio,” he said. “The AI system generates all sorts of candidates… the robotic lab tests them… and then it starts again.”

 

Schmidt, who was CEO and then executive chairman of Google for nearly two decades, has long been a prominent voice in the global AI debate. He has also played a significant role in shaping Silicon Valley, and has been involved in US national AI policy.

 

Is AI underhyped or overhyped?

Despite AI being in the news everyday, Schmidt maintains that AI was "underhyped, not overhyped."

 

“In this room everybody understands the ChatGPT moment... now 4.0 with a new one coming, Gemini… Claude 3 is the best one for programming… DeepSeek from China... Grok 3 from Elon Musk's XAI—all in the same equivalence class,” he said, pointing out that these tools are already deeply embedded in daily life, yet many people still fail to grasp the scale of what’s happening.

 

“You think of those as language-to-language… ask it a question… someone told me they use it for relationship advice… another used it for psychological advice,” he said. "Despite not being trained, tested, or vetted for such sensitive tasks, they are being used anyway."

 

Schmidt emphasised that people are unknowingly relying on AI in ways that far exceed its original design, and this quiet, unacknowledged adoption signals a future that is moving faster than most realise.

 

'AI programmers will replace programmers'

In the same conversation, Schmidt predicted a dramatic reshaping of the job market, saying, “In the next one year, the vast majority of programmers will be replaced by AI programmers.”

 

He also claimed AI would soon surpass top-tier human talent in fields like mathematics, leading to “superintelligence - computers that are smarter than the sum of humans.”

 

Schmidt, however, highlighted that AI is still a tool and humans must remain in control. “The scientists are in charge and AI is helping them, which is the correct order,” he said, adding that AI still required high-quality human-generated data.

 

“We don’t have enough data,” he said, emphasising the need for open, reproducible, peer-reviewed datasets to power responsible innovation.

 

Govt, human oversight needed in AI 

Schmidt warned of the national security risks posed by unregulated open-source AI models. “There’s lots of evidence that the new models, if they're unconstrained, can produce bad pathogens… just take an existing virus and modify it a bit and off we go,” he said.

 

Schmidt also reiterated that like most technologies, AI is also ahead of the law and is developing faster than anyone can keep up with. 

 

To mitigate such threats, he called for strong government involvement in AI research funding and oversight, defending the longstanding ecosystem in which public support powers university-led innovation. “Private philanthropy… you can’t make up the billions that the government provides,” he said, arguing that open-source AI tools should remain accessible to ensure academic institutions aren’t left behind.

 

Responding to a question from the audience, Schmidt stressed the importance of human oversight in AI systems: “The fact that AI is good at something doesn't mean the process that it's embedded in makes any sense.”