(Yicai) June 4 -- Around five more years are needed for physical artificial intelligence to achieve a ChatGPT-style breakthrough, while the number of robots worldwide may eventually surpass the human population, according to Zhang Yaqin, chair professor of AI Science at Tsinghua University.

Autonomous driving is poised to become the first commercialized application of physical AI, thanks to mature underlying technologies, with subsequent focus falling on large-scale commercial operation and the rollout of supporting industrial policies and regulations, Zhang, who is also dean of Tsinghua University's Institute for AI Industry Research, said in a recent exclusive interview with Yicai.

However, household robots face substantial hurdles to widespread commercialization due to the extreme complexity of domestic application scenarios, he added.

AI can be categorized into digital intelligence, physical intelligence, and biological intelligence, Zhang noted. Mainstream AI deployment remains concentrated within digital intelligence and has yet to substantively penetrate the physical world, he pointed out.

When AI expands into the physical domain, relevant tech will be embedded in robots, unmanned aerial vehicles, self-driving autos, and critical infrastructure, Zhang said, noting that at that stage, AI will be fully integrated into people's daily lives alongside a surge in application scale and industrial influence.

The commercialization of AI within biological systems will require an even longer development cycle, Zhang stressed. Ongoing research into biological AI spans brain-computer interfaces and innovative drug development, focusing on building foundational models for cells and proteins, with the ultimate goal of developing a universal foundational model for life science scenarios, he pointed out.

Driven by established core foundational technologies, digital AI applications have seen rapid growth over recent years, Zhang noted, predicting an explosive growth for physical and biological AI once their respective industrial infrastructure matures, accompanied by far-reaching socioeconomic impacts worldwide.

Regarding prevalent geopolitical risks, Chinese tech firms should target overseas expansion to adopt win-win collaboration and "AI for Good" as their core guiding principle, he said. "Companies should prioritize tackling shared global challenges, including public health crises, global warming, and population aging, with these areas having limited geopolitical exposure and closely aligning with the AI for Good philosophy."

Companies must invest in nurturing top-tier talent oriented toward the AI for Better initiative, ally with governments, non-governmental organizations, and charitable foundations to launch large joint projects addressing difficulties facing vulnerable groups, while also staying highly vigilant against inherent risks stemming from AI tech, Zhang stressed.

AI-related risks can be broken down into three core categories, Zhang said. First is the risk of technical malfunction, where, if multi-agent systems or robotic equipment gain access to critical physical infrastructure, including financial networks, nuclear power stations, and major power grids, unregulated technical failures could trigger catastrophic consequences, he noted.

The other two categories are misuse of AI tech and latent systemic social risks triggered by technological disruption, he added.

Chinese AI developers can secure solid overseas competitive edges by complying with local laws and cultural norms and embedding themselves into local industrial ecosystems bolstered by complete supply chain advantages, market-oriented operational capabilities, and an increasing number of original technologies, according to Zhang.

Overseas expansion does not equate to directly exporting products without modification, he said, noting that thorough localization, complete respect for local legislation and cultural customs, deep integration with local firms, and win-win ecosystem cooperation are indispensable prerequisites.

For example, stringent cross-border rules govern privacy protection for medical data across jurisdictions, so Chinese AI firms running foreign businesses must forge deep ties with local hospitals instead of assuming proprietary algorithms alone can resolve all challenges, he said.

Editors: Tang Shihua, Martin Kadiev