Artificial intelligence not all doom and gloom for jobs

Editor's note: As robotics and artificial intelligence technology continue to advance, people across various professions are asking a critical question: When will these technological changes affect their fields and in what ways? Yangcheng Evening News spoke to Liu Ning, a professor of intelligent robotics technology at Jinan University in Guangdong province, to explore the answers. Below are excerpts of the interview. The views don't necessarily represent those of China Daily.

As job opportunities in the digital sector continue to shrink, even fields such as computer science — once regarded as lucrative and highly coveted — have seen a sharp decline in starting salaries for recent graduates.

Companies today often employ an expert senior programmer, pair him or her with a junior assistant, and then invest in a software solution. AI-generated code is relatively standardized, which reduces the human role to reviewing and verifying.

In collaborative projects with hospitals, diagnostic systems that integrate medical imaging, patient records and lab results have, in many cases, outperformed general practitioners.

Yet, job contraction is not the whole story. While jobs involving simple, repetitive tasks are decreasing, new roles — involving skilled work such as data annotation and collection for large-scale model training and industrial imaging; on-site commissioning and maintenance of robotic systems; and specialized assistants in consulting fields such as law, insurance and healthcare who operate AI agents — are emerging.

For general-purpose humanoid robots to truly enter factory floors, multiple hurdles, ranging from data to hardware, are yet to be crossed. Much of what the public currently sees is demonstrative or performative, while actual industrial deployment is still limited.

The first bottleneck is the "brain". Today's large models are fed primarily with internet data, making them effectively "internet brains" that perform well at common-sense reasoning but lack real-world industrial operation data. Take defect detection in assembly, for instance. An untrained worker can instantly recognize a missing screw in a product, while a machine can only rely on predefined rules to check each condition.

The second bottleneck lies in the "hands". Human fingers are remarkably dexterous, integrating tactile, slip, force and temperature sensing in a sophisticated manner. Current robotic grippers, however, struggle to achieve such multi-sensory fusion. Their joint count is only a scaled-down version of the human hand, preventing them from performing delicate or precise movements. Actuation also remains a challenge.

The third bottleneck lies in the "foot". When a person jumps from a height of one meter, the impact upon landing can be several times their body weight. But human joints can absorb that impact. Existing robot motors and mechanical structures, however, are not built to withstand such hard impacts.

Manufacturing companies, eager to protect their competitive edge, keep production data strictly confidential, hindering cross-enterprise data sharing. Even with trimmed-down domain-specific models, training outcomes remain limited when data are confined to isolated silos.

Despite these challenges, it would be incorrect to conclude that robots are achieving little on factory floors. Robots can be categorized into three generations. The first generation operates on rules-based programming and has already matured in applications such as automotive welding and material handling.

The second generation relies on machine learning and excels at tasks that are difficult to define with explicit rules, such as grading apples on the basis of appearance. The third generation of robots is based on large-scale models and general-purpose intelligence. The pragmatic approach is to move forward in small steps with second-generation technology.

Looking at the global landscape, Silicon Valley in the United States leads in cutting-edge innovation, Germany excels in precision manufacturing and the production of drive components, and China stands out with its industrial scale and comprehensive supply chain.