Clinicians answer the AI cure-for-cancer claim with more than 100 brain tumour types

The promise has a definition problem before it has a scheduling problem. Cancer is a catch-all term for the uncontrollable growth of abnormal cells, usually named for the part of the body where it is found, and the diseases under that label differ in the cells they affect and in the genetic changes behind them ^[5]. Felix Sahm, a neuropathologist at University Hospital Heidelberg, told the Guardian that "there are more than 100 types of tumours in the brain, much more than in any other single organ" ^[7]. Sahm also said there are "so many unresolved questions in brain tumour research and brain tumour care" ^[8]. The Guardian's own verdict is that a single cure is "not only unlikely but nigh on impossible" ^[6].

Look at one cancer. Pancreatic cancer is the 12th most common cancer worldwide and the sixth most common cause of cancer death ^[10]. It ranks six places higher for deaths than for frequency ^[13]. The World Health Organization calls it "one of the cancer types with the least favourable prognosis" ^[11]. The gains of the past century arrived one modality at a time: chemotherapy and radiation made some solid tumours curable, and monoclonal antibodies allowed more direct targeting of abnormal cells ^[16].

The clinicians describe work at named points in a pathway. "AI can improve several parts of that process by helping us identify risk, detect disease, characterise tumours, select treatments and monitor," said Ajit Goenka, a radiologist and nuclear medicine specialist at the Mayo Clinic in Minnesota ^[9]. Loi and the other experts the Guardian interviewed expect efficiencies in treatment, diagnostics and pathology services, and more personalised and more equitable care ^[4]. Some of this is already clinical, including models reading lung imagery for early signs of malignancy and systems assisting the removal of brain tumours ^[12].

Pitched to funders, the claim is one sentence: AI cures cancer in five to 10 years, so the health line gets funded as a moonshot ^[1]. The pitch drops the unit of work, and it sits alongside claims that AI is responsible for human extinction before 2030 ^[2]. The Guardian reports those proclamations and does not say which executives made them ^[15]. A forecast that names no tumour type, no endpoint and no trial cannot be checked against anything in a planning cycle.

"Until AI can actually think and do experiments itself at scale, I'm not sure that we're going to find a cure for all cancer or medical illness in the next 10 years," said Sherene Loi, a medical oncologist at the Peter MacCallum Cancer Centre in Melbourne ^[3]. She set a condition, and she also said "Human disease is very complex" ^[14]. For a budget written this quarter, the fundable item is a detection or pathology step that reports a number inside the cycle; the cure claim will not resolve before several more planning rounds have closed.