Silicon Valley has no shortage of AI optimists when it comes to pie-in-the-sky predictions. That’s especially true in healthcare.
Earlier this year, Google DeepMind CEO Demis Hassabis told CBS News program “60 Minutes” that artificial intelligence (AI) could cure all diseases one day. Dario Amodei, CEO of Anthropic, has said he believes AI could one day eliminate most cancers.
Is AI finally the tool that will cure cancer, or is this another case of Silicon Valley hype getting ahead of science?
Experts agree that AI is reshaping how cancer is detected, studied and treated. But many caution that curing “most cancers” is a far more complex problem than it appears.
According to a recent blog post by the Cancer Research Institute, AI is already helping to process vast stores of medical knowledge, identify risk factors, aid in early diagnosis and accelerate drug discovery.
The institute cited an AI system that was able to predict the likelihood of pancreatic cancer from unrelated health records, rivaling the accuracy of expensive genetic tests.
Advertisement: Scroll to Continue
At Penn Medicine, researchers developed a tool that can detect cancer cells that are “easy to miss, or even invisible” to the eye, the Institute said. Artificial intelligence can also have a “crucial” impact on cancer treatment plans, such as by optimizing radiation doses, helping in surgeries and other actions. In research, AI can make it much more cost effective to discover drugs.
While these are welcome developments, experts emphasize that AI is a tool, not a magic bullet.
“The idea that AI will eliminate cancer is incredibly optimistic, but honestly, it is not as straightforward as it sounds,” said Kiara DeWitt, CEO of Injectco and a nurse. “Although AI can drastically improve diagnostics … a full cure is another matter entirely.”
While AI-driven image recognition software potentially boosts five-year survival rates by 30%, DeWitt added, “curing” every case is “still beyond AI’s reach alone.”
“Most cancer ‘cures’ are heavily individualized, dependent on patient genetics, immune responses and countless environmental variables no AI model fully comprehends today,” DeWitt said.
What’s more, the medical industry tends to oversell the capabilities of new technologies like AI, especially to investors. “New treatments still require human-led clinical trials, real-world effectiveness studies and regulatory approval, taking five to 10 additional years at minimum before reaching the market,” DeWitt said.
Dr. Maybell Nieves, a surgical oncologist and breast surgeon who writes for AlynMD, put it more bluntly: “I don’t see how the AI is going to be able to cure cancer,” she told PYMNTS.
“Decisions we make on the treatment of an oncologic patient are made with a multidisciplinary team and individualizing every patient,” Nieves said. “Sometimes medicine is not as mathematical as it might seem so I’m not sure that an AI will be ready to interpret every patients’ needs.”
See also: Exclusive: Color CEO Says AI-for-Oncology Copilots Detect and Treat Cancer Earlier
Cancer Is Not a Single Disease
Wyatt Mayham, CEO of Northwest AI Consulting, pointed out that cancer is not a single disease, but “hundreds of distinct genetic diseases” with wildly different behaviors, mutations and treatment responses.
However, Amodei’s optimism is “not entirely misplaced,” Mayham told PYMNTS. Its true potential in health care lies in three areas: Radically early and accurate diagnoses, shortening drug discovery timelines, offering hyper-personalized treatment.
For example, the Perception tool from the National Cancer Institute can analyze a tumor’s genomic and molecular data to predict which drug or combination of therapies will be most effective for an individual.
“The goal shifts from finding a universal cure to designing a unique cure for every single patient,” Mayham said.
Jeremy Gurewitz, CEO of Solace Health, pointed to additional complications, including ethical and socioeconomic barriers.
While AI has improved cancer detection and survival rates, such as a 9% improvement in mammography accuracy, only about 30% of cancers have a well-defined mutation that can be directly targeted by a drug.
Moreover, “data biases reduce AI accuracy in 25% of diverse populations,” Gurewitz told PYMNTS, citing 2024 JAMA data.
This means that models trained on narrow datasets may deliver inaccurate recommendations for underrepresented groups. Gurewitz called for more inclusive datasets and clinician oversight to ensure equitable care.
Ronen Cohen, vice president of strategy at Duality Technologies, added that unlocking AI’s full potential in medicine will require better solutions for privacy and data access. Much of the most valuable data for cancer research, such as clinical records, remains behind corporate firewalls.
“Unlocking these troves of data will take more than just smart algorithms; it will require strong, enforceable guarantees around privacy and security,” Cohen told PYMNTS. “AI companies are going to have to embed solutions like privacy-enhancing technologies into their offerings.”
With many hurdles to overcome, the institute said, the fight to cure cancer “is far from over.”
Read more:
AI System Can Predict Cancer Survival Prognosis Better Than Doctors, Researchers Say
SandboxAQ Launches Dataset for Training AI Models in Drug Discovery
Most Americans Take Cautious Approach to Generative AI in Healthcare
For all PYMNTS AI coverage, subscribe to the daily AI Newsletter.
