TikTok CEO Predicts AI Scientific Breakthroughs Within 3 Years

The Claim

Speaking at the DBS Asia Future Forward Conference, TikTok’s CEO made a notably specific prediction: AI will drive major scientific breakthroughs in physics, biotech, chemistry, and material science within the next two to three years.

“It’s very, very, very possible,” Chew said — which, for a CEO, is about as close to a pinky promise as it gets.

He framed it with a line that’s worth sitting with: “In the short term we are probably overestimating it, but in the long term we are probably underestimating it.” That’s not a hedge. That’s a map.

ByteDance Is Already Moving

This isn’t just conference-stage optimism. ByteDance’s Seed research team has already shipped Protenix-v2, an AI model built for biomolecular structure prediction and design — with direct applications in drug discovery.

That puts ByteDance in the same arena as Google DeepMind’s AlphaFold work, competing for relevance in a space where the stakes are genuinely high. Last week, a separate $1.8 billion initiative involving the U.S. government, Meta, and Google was announced to develop AI models that predict cellular behavior — another signal that scientific AI is attracting serious capital and serious players.

ByteDance is also exploring collaborations with scientists directly. “There’s an effort to see how we can collaborate with a lot of scientists to make some progress in science,” Chew said.

The Alignment Problem Isn’t Going Away

Chew didn’t just sell the upside. He flagged two real concerns worth tracking:

• AI alignment — as models take on more autonomous tasks, ensuring they operate within human ethical boundaries becomes harder and less optional.
• Cybersecurity — AI is dramatically lowering the cost for bad actors to run fraud and attacks at scale. Platforms like TikTok are already feeling this pressure.

He gave a grounded example: AI agents with purchase permissions might buy far more than a user intended, simply because they’re optimized to complete the task. It’s a small illustration of a large problem — autonomous systems that are helpful in narrow ways can be costly in unexpected ones.

What This Means for AI Tool Watchers

A few things stand out here for anyone tracking the AI tools ecosystem:

• Scientific AI is becoming a product category. Biomolecular modeling, drug discovery pipelines, and materials research tools are moving from research labs toward commercial platforms. That’s a space worth watching.
• Big tech’s AI pivot toward science is accelerating. When ByteDance, Google, Meta, and the U.S. government are all funding the same general direction, the tooling ecosystem around it tends to follow quickly.
• The cybersecurity angle is underrated. As AI agents gain more permissions and autonomy, the tools built to audit, monitor, and constrain them will matter as much as the agents themselves.

Chew’s timeline — two to three years — is aggressive. But the infrastructure being built right now suggests the direction is correct, even if the schedule slips. The smarter move is to watch which tools are already operating in these spaces, not wait for the breakthrough announcement.