What happened

Shou Zi Chew, the CEO of TikTok, said artificial intelligence could produce a major scientific breakthrough within the next two or three years. He spoke at the DBS Asia Future Forward Conference in Singapore. He named physics, biotechnology, chemistry, and materials science as possible areas.

This is a forecast, not a promised discovery. Chew did not name a specific experiment, product, or date. The useful question is therefore not whether AI will magically solve science. It is whether current progress can help researchers find and test ideas faster.

The research behind the forecast

TikTok's parent company, ByteDance, is expanding AI work beyond consumer applications. Its Seed research team unveiled Protenix-v2 in April. The model focuses on predicting and designing biomolecular structures. That work could support drug discovery, according to the report. It does not mean that Protenix-v2 has already produced a new medicine.

ByteDance is also looking for ways to work with scientists. The report places this effort inside a wider industry push. The U.S. government and two large technology companies, Meta and Google, recently announced participation in a $1.8 billion initiative. The goal is to develop AI models that predict how cells behave. Such models are intended to speed biological research and drug discovery.

Why it matters

Scientific research often depends on exploring many possible explanations. AI could become useful for organizing evidence, suggesting designs, or narrowing the options researchers examine. Those are reasonable implications of the projects described. They are not proof that a breakthrough will arrive on schedule.

Chew's view is deliberately balanced. He said AI's short-term impact may be overestimated, while its long-term potential may be underestimated. That distinction matters because a bold forecast can attract attention before researchers have measured real results. A discovery must still be supported by clear research evidence.

The safety question

Chew paired his optimism with a warning. More capable AI systems will need stronger safeguards, especially when they can act on their own. Users might give an AI agent broad permissions without understanding what could happen.

He used autonomous purchases as a simple example. An agent trying to complete a task might buy more than the user intended. He also warned that better AI could make fraud cheaper for malicious actors. Companies would then need better tools to defend their platforms and customers.

This is a system-level concern. The issue is not only what an AI model can say. It is also what the model can do when connected to accounts, money, or other services.

What is known, and what is not

The confirmed facts are limited but clear. A TikTok executive made the forecast. ByteDance has described research connected to biomolecular design. Major technology and public-sector groups are investing in AI for biology. Chew also said safeguards must improve.

The unknowns are larger. The report does not identify the breakthrough, the field most likely to produce it, or the evidence needed to call it major. It also does not show that the three-year prediction has been independently validated.

What to watch next

The next useful signals will be reproducible research results, public evidence from scientist collaborations, and clearer safety controls for autonomous systems. If AI produces important discoveries, researchers should be able to explain the evidence and repeat the result. If AI agents gain more permissions, users and companies will need ways to limit, review, and stop their actions.

The three-year timeline is a headline. The real test will be measurable science and safer deployment.

Source: CNBC