Could artificial intelligence drive major scientific breakthroughs within the next 3-years? TikTok CEO Shou Zi Chew believes it could, anticipating that advances in AI will advance discoveries in physics, biotechnology, chemistry, and materials science. Speaking at the DBS Asia Future Forward Conference in Singapore on October 12, Chew outlined the technology’s developing potential while warning about cybersecurity and AI safety risks.
AI Could Accelerate Scientific Discovery
Chew anticipates AI to transform scientific research beyond its recent applications in content generation, recommendation systems, and automation.
“In the next 2 to 3 years, we’re going to have a real scientific breakthrough: physics, biotech, chemistry, material science, as of AI,” Chew stated.
His prediction shows increasing investment in AI-powered scientific research. Technology companies and research institutions are developing models capable to anticipating molecular structures, simulating biological systems, and detecting promising materials.
ByteDance, TikTok’s parent company, is participating in this expansion. In April, its Seed research team launched Protenix-v2, an AI model designed for biomolecular structure prediction and design. The technology ought to guide drug discovery via supporting researchers better understand molecular interactions.
Chew also confirmed that ByteDance is exploring collaborations with scientists to advance AI-driven research.
Scientific AI Gains Momentum Across Industries
The developing interest in scientific AI broadens beyond ByteDance. Major technology companies and government organizations are making an investment in heavily foundation models designed for scientific applications. Current venture include a reported $1.8 billion effort involving the U.S. Government, Meta, and Google to develop AI systems that are expecting cellular behavior.
Moreover, scientific progress needs more than right predictions. Researchers must validate model outputs against experimental evidence, account for uncertainty, and ensure that systems generalize beyond their training data.
Chew stated that expectations surrounding AI may not align with its real development timeline.
“I think in the short term we’re likely overestimating it, however within the long term we are likely underestimating it,” he stated.
AI Agents Launch New Security Challenges
Despite his optimism, Chew warned that increasingly autonomous AI systems launch substantial cybersecurity risks.
AI agents can execute tasks, access sensitive information, and interact with external services. Moreover, granting these systems extensive permissions without right safeguards creates opportunities for unplanned actions.
Chew emphasized autonomous purchasing as one example. An AI agent tasked with completing a transaction might spend more than supposed if its permissions and spending limits are poorly described.
He also warned that sophisticated AI models could decrease the cost of fraudulent activities, making cyberattacks more accessible.
These concerns underscore the significance of permission management, human oversight, and persist safety evaluation as organizations deploy increasingly capable AI systems.
