TikTok CEO Predicts AI Breakthroughs in Science Within Three Years

TikTok CEO Shou Zi Chew believes artificial intelligence is about to break out of the group chat and into the laboratory. Speaking at the DBS Asia Future Forward Conference in Singapore in October 2026, Chew predicted that AI breakthroughs in physics, biotech, chemistry and material science would arrive within the next two to three years, according to CNBC.

"In the next two to three years, we're going to have a real scientific breakthrough: physics, biotech, chemistry, material science, because of AI," he said. Chew framed the prediction as part of a pattern he keeps describing: the short-term buzz around AI usually runs ahead of reality, while the long-term potential keeps getting underestimated.

Chew's forecast also fits his own company's direction. In April 2026, ByteDance's Seed research team unveiled Protenix-v2, an AI model for biomolecular structure prediction and design with potential applications in drug discovery. Chew said ByteDance is now exploring ways to collaborate directly with scientists, and described the early work in this area as very exciting.

Why physics and biotech lead the AI breakthroughs race

Physics and biotechnology look like an odd pair, but they share a problem: both are bottlenecked by costs that AI is unusually good at cutting. Physics research, from materials science to particle physics, is often limited by the sheer computational cost of simulating complex systems. Biotech, by contrast, is usually limited by the slow and expensive grind of real-world experiments: growing cells, running trials and waiting for biology to play out in a dish.

AI attacks both bottlenecks from opposite ends. On the physics side it offers faster and cheaper simulation and modeling. On the biotech side it speeds up hypothesis generation and increasingly steers automated lab equipment. For anyone studying or eyeing a career in the sciences, that dual pressure is the point: the skills at the intersection of computation and bench work are where the next wave of AI breakthroughs — and the next decade of jobs — is likely to cluster. More coverage of the field lives on the AI news topic page.

AI breakthroughs are already reshaping medicine

Chew's timeline is aggressive, but it is not coming out of nowhere — AI has already produced at least one Nobel-level result. The 2024 Nobel Prize in Chemistry went to Demis Hassabis and John Jumper for AlphaFold's protein-structure prediction work, with David Baker sharing the honor for computational protein design, as announced by the Royal Swedish Academy of Sciences. That remains the clearest proof yet that AI breakthroughs in science can be real, peer-reviewed and prize-worthy.

ByteDance's own bet on AI breakthroughs in biomolecular design is part of a wider rush. Earlier this month, the United States government, Meta and Google joined an initiative worth nearly two billion dollars to build AI models that predict how cells behave, aimed at accelerating biological research and drug discovery. Drug discovery is the high-value payoff everyone keeps circling: it is one of the few categories big enough to justify the hundreds of billions of dollars in AI infrastructure spending now flowing through the industry.

The counterpoint: not every promised wave of AI breakthroughs lands on schedule

Before penciling Nobel ceremonies into your calendar, consider the skeptics' file. OpenAI released seven hundred twenty-two math manuscripts from an unreleased model earlier this year; when mathematicians ran the claims through the Lean formal-verification system, only about forty-two percent cleared the checks, prompting public pushback. The gap between a forecast and a verified result can be wide, even when the underlying model is genuinely impressive.

Chew's window also sits in a crowded field of executive timelines that do not agree with each other. Anthropic CEO Dario Amodei told the World Economic Forum in Davos in 2025 that AI could become smarter than a Nobel Prize winner within two to three years, and compress fifty to one hundred years of biology and medicine progress into five to ten. Google DeepMind CEO Demis Hassabis, whose lab actually won the chemistry Nobel, talks instead about a "new renaissance" in materials science, drug design and math over ten to fifteen years, as reported by Fortune in February 2026. Similar comparisons of the competing timelines have been reported by Tech Insider.

The more convincing pattern may be smaller and independently checkable — like Claude's recent bound on a Riemann problem, which shows AI doing real math without claiming to have solved the unsolvable. Incremental, verified results are arguably a better predictor of where a few years of AI-assisted science actually lands than a splashy announcement of hundreds of simultaneous results.

The alignment question Chew keeps raising

What sets Chew's version of the prediction apart is the caveat he keeps attaching to the coming AI breakthroughs. He asked how people would align increasingly powerful models with the things that are obvious to humans — the classic alignment problem, framed in plain language — and insisted that human judgment, creativity and authenticity still matter. He has previously called for careful, thoughtful guardrails that do not kill innovation.

He also flagged the unglamorous side of more capable AI: cybersecurity. Chew warned that people may hand AI agents broad permissions without understanding the consequences — the agents could make purchases users never fully intended — and that smarter models make fraud dramatically cheaper to pull off. The point lands for anyone who already lets an agent book, buy or schedule on their behalf.

What AI breakthroughs mean for your future

Here is the reader-impact version. If Chew's timeline is even roughly right, the medicine you use in your twenties and thirties — new drugs, new materials for everything from batteries to implants — may be designed with AI in the loop. The research careers most worth chasing are increasingly hybrid: biology plus machine learning, chemistry plus simulation. And the ethics framing matters too, because someone has to write the guardrails, run the verification and say no when a model overreaches — all plausible job descriptions by the time today's students graduate.

Even if the breakthroughs arrive later than his deadline, the direction of travel is hard to argue with. The money is committed, the models keep improving and the labs are hiring. Whether AI breakthroughs arrive on Chew's schedule or on Hassabis's slower one, the race toward verified, peer-reviewed AI science is the story worth watching — and it is only just getting started.