Arm CEO: AI Could Help Humanity Conquer Cancer in Our Lifetime, but Chip Supply Is the Biggest Bottleneck

Arm CEO Haas said in an interview with the BBC that AI has disruptive potential in healthcare applications, can shorten drug development and testing times, and may help humanity conquer cancer within this generation.

Article author, source: BBC

The Arm CEO stated that artificial intelligence might help humanity conquer cancer within our lifetime. However, before this goal is achieved, the entire industry first needs to solve a real-world problem: producing enough chips to provide the computing power that AI requires.

“I have always thought that the truly disruptive application area for AI is healthcare.” Arm CEO Rene Haas said in a BBC interview video released on Tuesday local time, “AI can not only shorten the time needed for drug development, but also the time required for drug testing.”

“First, more chip fabs are needed”

Haas joined Arm in 2013 and became CEO in February 2022, after working for Nvidia for seven years as Vice President of Computing Products. In the interview, he said that AI might solve problems that are very difficult for humans alone: “I believe that in our lifetime, AI will help humanity conquer cancer.”

But he pointed out that this will require massive computing power, and currently the chips needed for cancer research and humanoid robotics are being heavily consumed by AI companies planning to build data centers on the scale of several gigawatts. “At present, the supply of computing power is indeed very limited. Unless we can build data centers in space, the first thing we need is more chip fabrication plants, and I can tell you that unequivocally.”

Haas indicated that the scale of AI’s demand for chips and memory is at a level the semiconductor industry has never faced before, and that building new chip fabrication plants cannot solve the problem quickly—a single chip fab may require tens of billions of dollars in investment, and construction usually takes two to three years. He said the semiconductor industry is currently in an “absolutely supply-constrained environment,” and tight supply will persist for some time.

Conquering cancer is not just about computing power, and data center expansion raises opposition

Even if the chip bottleneck is resolved, this does not guarantee that AI will conquer cancer in the coming years. Medical researchers point out that cancer is not a single disease but rather a large group of over 100 different diseases, with significant variations among patients. While AI can indeed help identify potential drug targets and predict drug actions, there is still limited evidence that it can have a real impact on clinical medicine, and the vast number of variables in the human body makes it challenging for models to reliably predict the effectiveness of candidate drugs.

Other chip executives have made similar warnings. Micron Technology COO Manish Bhatia said in January this year that the current memory shortage is “truly unprecedented,” with HBM taking up a large amount of capacity, leading to “significant supply shortages” for traditional products such as mobile phones and PCs; Qualcomm CEO Cristiano Amon said in the February earnings call that “industry-wide memory shortages and price increases” will determine the scale of this year’s smartphone market.

In recent years, data center construction in the US has rapidly expanded, with more than 1,400 centers built or approved by the end of last year. As data centers continue to move into communities, their impact on air quality and water resources has also become contentious. A Gallup survey in May this year found that out of 1,000 US adults, 70% opposed building data centers for AI in their area, and nearly half expressed “strong opposition.”

","author":"ME精选","type":"转载"}