Nobel Laureate Michael Levitt: AI Is Making Innovation Cheaper, but Human Judgment Still Matters

SFC, 21st Century Business Herald Correspondent Li Yinong

At the ninth Hongqiao International Economic Forum, during the parallel session “Pudong and the World: Technological Innovation Driving Industrial Upgrading, Smart Integration Shaping a Shared Future,” Nobel laureate in Chemistry Michael Levitt shared his insights into how artificial intelligence is transforming scientific research, industrial development and global collaboration.

“AI essentially is able to accelerate everybody,” Levitt said. But as AI makes innovation faster and the cost of experimentation lower, he believes the more important question is no longer simply what technology can do, but what people should choose to do with it.

Drawing on more than six decades of experience in science and computing, Levitt highlighted three key dimensions of the AI revolution: the importance of human judgment and verification, AI's growing impact on industries such as biotechnology, and the need for closer collaboration between China and the United States.

The Hardest Part Is Knowing What Matters

Scientific discovery has always been a process of trial and error. Researchers develop hypotheses, test them and frequently discover that their initial assumptions are wrong. The challenge has traditionally been the time, money and resources required to learn from those mistakes.

AI is changing that equation.

With AI tools and agents, researchers can explore more possibilities, test hypotheses and identify errors at a fraction of the time and cost previously required. Tasks that once involved days of programming can sometimes be completed in minutes, allowing scientists to devote more time to the questions that matter.

“AI doing is cheaper. The question is what matters,” Levitt observes.

That shift, in his view, makes human curiosity and judgment more important than ever. As AI makes an increasing range of tasks possible, people need to become better at distinguishing between what can be done and what is actually worth doing.

“We can do anything, but it's critical to do the important things,” he says. “We need to learn how to ask what is worth solving.”

Levitt's own experience reflects how dramatically the relationship between humans and computers is changing. Having started writing computer programs in the 1970s, he began using AI tools such as ChatGPT and Claude in November 2022. Initially, he used them mainly to ask questions and generate answers. Today, he increasingly works with AI agents that can read files, write and modify programs, execute code and check their own results.

Yet greater efficiency does not necessarily mean greater reliability. AI systems can produce answers at extraordinary speed, but those answers may be incorrect or simply irrelevant.

As Levitt puts it, “AI makes wrong answers very quickly as well, or maybe not wrong answers but irrelevant answers.”

He believes verification will therefore become a critical part of the AI era. Referring to the success of AlphaGo, he points out that an important factor was the ability to check its answers quickly. The same principle, he argues, needs to be applied to new drugs, materials, chemical discoveries and other AI-generated solutions.

“We need to start focusing on verification schemes for new drugs, for new materials, for new chemistries,” he says. “The verification is going to be what tells us whether the AI is doing something that is worth doing versus not worth doing.”

There is also a question of timing. AI can process information and generate responses extremely quickly, while meaningful human judgment sometimes requires more time and reflection. As Levitt notes, “AI is thinking very quickly and sometimes human beings need to think more slowly.”

The challenge, then, is not to compete with AI on speed, but to find the right balance between machine efficiency and human judgment.

How AI Is Taking Science from the Lab to the Market

Beyond individual productivity, AI is beginning to reshape entire industries. Biotechnology and pharmaceutical research provide some of the clearest examples.

For decades, predicting how proteins fold into their three-dimensional structures was one of biology's most challenging problems. Levitt began working on protein folding in 1969 and conducted his first simulations in 1975. The field made gradual progress over several decades before a major breakthrough arrived with DeepMind's AlphaFold in 2020.

The system demonstrated remarkable capabilities in protein structure prediction, marking a turning point in computational biology. A challenge that had occupied scientists for roughly half a century could now be addressed with unprecedented speed and accuracy.

The implications extend well beyond academic research. Understanding protein structures can help scientists investigate biological mechanisms, identify potential drug targets and develop new medicines.

China is also emerging as an important testing ground for the industrial application of AI in drug discovery. Insilico Medicine, a biotechnology company based in Shanghai's Pudong New Area, provides a notable example.

Its work on a potential treatment for pulmonary fibrosis demonstrated how AI could accelerate early-stage drug development. The company moved from target identification to a preclinical drug candidate in less than 30 months, illustrating the potential to compress processes that traditionally take considerably longer.

AI is also creating opportunities in medical imaging. Advanced MRI technologies, combined with insights from physics and AI, are enabling researchers to look beyond the anatomical structure of tumors and investigate their metabolic activity and molecular characteristics. Such developments could provide a more comprehensive understanding of disease and support new approaches to diagnosis and treatment.

These advances do not mean AI can eliminate the uncertainties of drug development. Clinical trials, regulatory review and scientific validation remain indispensable. However, by making early-stage research faster and more efficient, AI could help reduce development costs and allow researchers to explore a much wider range of potential treatments.

The broader shift is from using AI to improve individual tasks to integrating it throughout the innovation process. In biotechnology, that could fundamentally change the economics of scientific discovery.

China and the US Have More to Gain by Working Together

The AI revolution is global, and its future will be shaped not only by technological breakthroughs but also by how countries choose to work together.

Levitt sees considerable potential for closer cooperation between China and the United States, two countries with different but complementary strengths.

“China has been very quick in developing its own AI, even developing its own chips for AI,” he says. “China is very good in application.”

To illustrate the difference, Levitt draws on a familiar experience: using smartphones in the two countries.

“I always compare my Chinese phone to my American phone,” he says. “And I could not manage in China without a phone. I could probably manage in America without a phone.”

In China, he explains, a smartphone has become essential for everyday activities. “You can't get a taxi, you can't buy anything, you can't order in a restaurant.” In the United States, many of these activities can still be done in more traditional ways, although digital payments and QR-code ordering are becoming increasingly common.

The comparison offers a glimpse into two different approaches to technological development. The United States has a long tradition of fundamental scientific research and has made major contributions to the development of AI technologies. China, meanwhile, has demonstrated considerable strength in translating technological advances into practical applications and integrating digital services into everyday life.

For Levitt, the interaction between these two approaches will be particularly interesting to watch.

“It's going to be very interesting to see how these two different countries interact,” he says.

Scientific discovery has always depended on the exchange of knowledge and ideas across borders. Cooperation between Chinese and American researchers, universities and companies could help accelerate progress in areas ranging from biotechnology and medicine to fundamental science.

Healthcare is a particularly important area. Diseases do not respect national borders, and advances in medical research have the potential to benefit people around the world. Combining different research capabilities, scientific expertise and technological approaches could open new avenues for innovation.

Levitt is firmly in favor of collaboration. “I think this collaboration is going to be terribly important,” he says.

Ultimately, the value of AI should not be measured solely by how powerful the technology becomes. Its real significance lies in whether it can help humanity solve difficult problems, improve healthcare and expand the boundaries of scientific knowledge.

AI is making innovation faster and experimentation less costly. But technology alone cannot determine the direction of progress. Human judgment will remain essential, while international collaboration can help ensure that the benefits of AI reach far beyond national borders.