TikTok CEO Shou Zi Chew expects artificial intelligence to accelerate breakthroughs in physics, biotechnology, chemistry and materials science within the next two to three years, arguing that the technology’s long-term scientific potential remains underestimated even as companies and governments invest billions of dollars in its development.
Speaking at the DBS Asia Future Forward Conference in Singapore on Monday, Chew said the pace of AI progress was opening up possibilities beyond its current applications in consumer technology, content creation and digital services.
“In the next two to three years, we’re going to have a real scientific breakthrough: physics, biotech, chemistry, material science, because of AI. It’s very, very, very possible,” Chew said.
Register for the next Tekedia Mini-MBA.
Register for Tekedia AI in Business Masterclass.
Join Tekedia Capital Syndicate and co-invest in great global startups.
His forecast comes as technology companies increasingly direct AI research towards scientific problems, including drug discovery, biological modelling and the development of new materials. These applications could extend AI’s economic impact beyond productivity gains in existing industries, potentially changing how researchers investigate complex systems and develop new treatments or technologies.
However, Chew also warned that the growing capabilities of AI systems would raise difficult questions about control, cybersecurity, and the ability to ensure that autonomous machines act in ways consistent with human expectations.
“If that happens, then you know the question would be: how do you align the models? How do you align the machines to be doing things that are obvious to us as human beings?” he said.
The comments highlight a central tension in the AI industry: companies are racing to develop more capable systems while confronting the challenge of ensuring those systems remain secure, predictable, and aligned with human intentions.
Chew’s outlook comes as ByteDance, TikTok’s Chinese parent company, expands its AI research beyond the consumer-facing applications that have helped define its business.
In April, ByteDance’s Seed research team unveiled Protenix-v2, an AI model focused on biomolecular structure prediction and design. The work has potential applications in drug discovery, where understanding the structure and interactions of biological molecules can help researchers investigate diseases and develop treatments.
ByteDance is also exploring opportunities to work with scientists on research projects, suggesting that its AI ambitions extend beyond improving recommendation systems, content generation, and other digital services.
“There’s an effort to see how we can collaborate with a lot of scientists to make some progress in science as well,” Chew said. “There’s been some work that we’ve done in this area, and it is very exciting.”
The company’s research reflects a broader shift among major technology firms towards using AI to tackle problems that have traditionally required extensive laboratory work, specialized expertise and lengthy experimentation.
AI models can help researchers analyze large datasets, identify patterns, and generate predictions that guide further experiments. In areas such as molecular biology, these capabilities could help narrow the range of possibilities scientists need to test, potentially accelerating parts of the research process.
But scientific breakthroughs remain difficult to predict. AI-generated predictions must often be validated through experiments, and progress in a specialized task does not automatically translate into a successful drug, commercial technology, or practical discovery. Chew’s two-to-three-year timeline is therefore a forecast of what the technology could enable, rather than a guarantee that breakthroughs will emerge across all the fields he identified.
The investment behind the sector nevertheless suggests that scientific applications are becoming a major focus of AI development. Last week, the US government, Meta and Google announced their participation in a $1.8 billion initiative to develop AI models capable of predicting how cells behave, with the goal of accelerating biological research and drug discovery.
Such efforts illustrate how AI development is moving beyond chatbots and consumer applications towards systems designed to model complex biological processes. They also signal growing interest in combining computing resources, scientific datasets and research expertise to address problems that individual laboratories may struggle to solve at scale.
Long-Term Potential Meets Immediate Security Concerns
Despite his optimism about AI’s scientific impact, Chew argued that the technology’s near-term capabilities may be receiving more attention than its longer-term implications.
“The trend is very very clear. It’s that we are only at the beginning of what artificial intelligence can probably do,” he said, pointing to what he described as a steep curve of progress, particularly in science.
“I think in the short term we are probably overestimating it, but in the long term we are probably underestimating it,” he added.
The distinction captures the uncertainty surrounding the AI investment boom. Companies are spending heavily on computing infrastructure, model development and new applications, but the pace at which those investments translate into reliable products and measurable economic returns remains an open question.
Scientific research could become one of the more consequential tests of AI’s value. Breakthroughs in drug discovery, materials science or chemistry could have effects well beyond the technology sector, although the scale and timing of those benefits will depend on whether model outputs can be translated into validated discoveries.
For Chew, the opportunities must be considered alongside growing security risks. He warned that AI agents capable of taking actions on behalf of users could create vulnerabilities if people grant them extensive permissions without fully understanding the consequences.
One concern is that an agent authorized to complete a task could make purchases or take other actions beyond what a user intended. Systems designed to pursue a goal autonomously may carry out instructions in ways that technically satisfy the request but exceed the user’s expectations, particularly where spending limits and approval requirements are not clearly defined.
The risks extend beyond consumer transactions. As AI models become more capable, malicious actors may be able to use them to carry out fraudulent activities more cheaply and at greater scale, increasing the burden on companies responsible for protecting users and digital platforms.
“I think the cybersecurity risks are quite big,” Chew said.
He warned that companies would need stronger defensive capabilities as the cost of generating security threats falls.
“So you need the right tools to now you know defend against the fast falling cost of creating problems for platforms like us, security problems for us,” he said.
These concerns are relevant for TikTok because AI is integrated into the platform’s operations, including content creation, recommendations, search and moderation. More capable systems could improve user experiences and operational efficiency, but they could also create new avenues for abuse and raise the complexity of detecting malicious activity.
Chew’s remarks ultimately point to two sides of the same technological shift. AI could help researchers make discoveries that reshape industries, but its growing ability to reason, generate content, and act autonomously will also demand more robust safeguards.