TikTok CEO Shou Chew predicts AI science breakthroughs within three years
TikTok chief executive Shou Chew says artificial intelligence will drive major scientific advances in physics and biotech within two to three years.
Key points
- Shou Chew expects AI to enable breakthroughs in physics, biotech, chemistry and materials science within two to three years.
- Chew warns that users may grant AI agents excessive permissions, creating real cybersecurity risks in the short term.
- ByteDance is developing internal AI models for drug discovery and biomolecular structure prediction.
TikTok chief executive Shou Chew stated on Monday that artificial intelligence will likely produce a major scientific breakthrough within the next two to three years. He made these remarks at DBS Bank’s Asia Future Forward Conference in Singapore, according to a report by CNBC.
Chew identified physics, biotechnology, chemistry and materials science as the fields most likely to see rapid advancement due to AI capabilities. He expressed strong confidence in this timeline, noting that such developments are very possible, CNBC reported.
Context and Background
Shee Tse Koon of DBS moderated the session, according to the conference programme. While predicting long-term gains, Chew cautioned that society may currently overestimate AI’s immediate capabilities. He suggested that the technology still lacks the nuance required for certain complex tasks, Bloomberg reported.
He specifically noted that human brains still outperform AI models in visual recognition and long-term memory retrieval. Chew emphasized that current technology has not yet reached the level of human cognitive complexity in these areas, according to Bloomberg.
Who Is Affected
Chew highlighted significant cybersecurity risks associated with AI adoption, describing them as quite real in the short run. He warned that users might grant AI agents broad permissions without fully understanding the potential consequences of those actions, CNBC reported.
He illustrated this risk with an example of an agent authorized to make purchases. Such an agent might buy more items than intended in its eagerness to complete a task efficiently. Stronger AI models also lower the cost for bad actors to attack platforms like TikTok, Chew said.
ByteDance, TikTok’s parent company, is actively developing its own AI models to address these challenges and opportunities. In April, its Seed research team published Protenix-v2, a model designed for biomolecular structure prediction and design, according to the source material.
The company also operates an AI drug discovery unit called Anew Labs. Chew indicated that ByteDance is exploring ways to collaborate with scientists to accelerate progress in scientific fields, similar to recent efforts by the US government, Meta and Google.
What should you do?
Security professionals should monitor how AI agents are being integrated into organizational workflows and review permission scopes carefully. Understanding the limitations of current AI models in visual recognition and memory can help mitigate over-reliance on automated systems.
What to do and how to stay safe: TikTok
- Review and restrict permissions granted to AI agents to prevent unauthorized actions like excessive purchasing.
- Monitor for increased attack volumes on platforms, as stronger AI models reduce the cost for bad actors.
- Evaluate internal AI tools for limitations in visual recognition and long-term memory retrieval compared to human performance.
- Stay informed about vendor updates regarding AI security risks, as no specific fix for these systemic issues has been confirmed yet.
Step-by-step guide: Prompt Injection Warning Signs: How to Spot AI Manipulation
General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.
Frequently asked questions
When does Shou Chew expect AI to deliver scientific breakthroughs?
Chew expects major breakthroughs in fields like physics and biotech within the next two to three years.
What specific cybersecurity risks did Chew mention?
He warned that users may give AI agents wide permissions without understanding consequences, leading to unintended actions.
Is ByteDance developing its own AI models?
Yes, ByteDance develops its own models, including Protenix-v2 for biomolecular structure prediction and an AI drug discovery unit.
