Progress and limitations of AI in cancer diagnosis: Experts point out that there is no universal antidote

2026/09/29 14:15

Artificial intelligence (AI) can help analyze computed tomography (CT) images, and researchers have even found that AI may be able to find subtle clues related to future pancreatic cancer from images that were judged to be normal in the past. (Drawing by this newspaper)

[Compiled by Chen Chengliang/Comprehensive Report] Can artificial intelligence (AI) really find an antidote to cancer for humans in just 10 years, as the technology giants say? At a time when major technology companies are proposing ambitious medical blueprints, clinical oncology experts remind that the complexity of biomedicine cannot be solved by computing power alone.

The British "Guardian" reported that Anthropic CEO Dario Amodei has said that cancer may be cured in the next 10 years; OpenAI CEO Sam Altman also believes that as long as it has enough computing power, AI may be able to find a cure for cancer; a team under Google DeepMind has even listed "solving all diseases" as its goal.

Please read on...

However, Peter. Sherene Loi, a breast cancer oncologist at the Peter MacCallum Cancer Center, points out that the reality is not that simple. Cancer is not a single disease, but a large group of diseases involving changes in different cells and genes.

Felix Sahm, a neuropathologist at the University Hospital Heidelberg in Germany, also pointed out that there are more than 100 types of brain tumors alone, and the tumor characteristics of different cancers and patients may also be different, making it more difficult to find universal treatments.

Old images hide clues, AI diagnosis shows results

Experts pointed out that the current more specific progress of AI is not to directly develop antidotes, but to assist in early detection and cancer diagnosis. A study by the Mayo Clinic in the United States shows that a machine learning model can find subtle signals that may be related to future pancreatic tumors from old computed tomography (CT) images of patients that were judged to be normal a few years ago.

Emilia Javorsky, a scientist at the Future of Life Institute, emphasized that biological problems cannot be solved by computing power alone. Even if AI can quickly propose a large number of candidate drug hypotheses, they still need to undergo laboratory verification and human clinical trials. In the face of this long research, AI is currently more like a tool for the medical community to speed up the search for clues, rather than a panacea key to the antidote to cancer.