“AI is supposed to change all that. But so far it hasn’t.”
Our reading, the sentence that was checked:
So far, AI has not changed the toll that cancer takes on people.
Assessment: Unresolved Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, F Cancer, paragraph 2.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-f-cancer-1e2f0f-a1
We recorded 7 claims from this post. See every claim from this post.
The wording above is the version this site observed on 2026-10-28. The source may have been edited after 2026-03-16. This page does not claim the wording was captured then.
What the claim commits to
- Kind
- Observation
- Deadline, as the speaker wrote it
- None stated. This site does not add one.
- Deadline state when assessed
- No horizon
- Deadline state on 2026-10-29
- No horizon (the release’s reference date)
Conditions
- so far (qualifier, stated by the speaker)
From the speaker
- AI
- change all that
- so far it hasn’t
Our interpretation
- The ordinary reading is that AI had not materially reduced cancer deaths or suffering by publication.
- The surrounding discussion concerns bereavement and cancer's human toll. It does not clearly distinguish population mortality from benefits to particular patients.
- “So far” describes cumulative experience through 2026-03-16. No starting date is supplied, so the target period has an unspecified beginning.
- Better diagnostic accuracy alone does not establish reduced mortality or suffering. Clinical benefits must also be attributable to an intervention involving AI.
Counts as supported if: Evidence with adequate coverage of clinical and population outcomes establishes that AI had produced no material reduction in cancer deaths or suffering through publication.
Counts as contradicted if: Reliable comparative human evidence establishes a material reduction in cancer deaths or suffering attributable to AI through publication. Improvements in prediction alone are insufficient.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Was it true for the period described?
Outcome: Unresolved
The record was checked, and the evidence available by the cutoff does not settle it.
Scored on a plain reading. The speaker named no measure for this claim, so it was judged on the reading an informed reader would take, stated below under our interpretation. Both reviews had to find that reading fair.
- Evidence as of
- Deadline state when assessed
- No horizon
- Assessment recorded
- Policy version
- 2026-10-28.7
Rationale
The evidence does not establish the broad negative claim. SHIELD-RT reports fewer acute-care visits with an AI-guided care strategy, and MASAI reports improved screening outcomes. Those findings make a literal zero-benefit reading doubtful. But SHIELD-RT does not isolate AI's incremental contribution, and MASAI does not establish reduced mortality or suffering. Official death counts cannot resolve attribution. A firm verdict would depend too heavily on whether Marcus meant population-level transformation or any meaningful patient benefit.
Limitations
- The claim gives no magnitude, geography, or precise outcome.
- Screening outcomes and acute-care use are not interchangeable with cancer mortality.
- The search cannot establish a universal absence of benefit.
Evidence (3 items)
Context only: Journal of Clinical Oncology, indexed by PubMed,
reduced acute care rates versus standard of care.
Why it matters here: Among patients selected as high risk by machine learning, additional evaluations reduced acute-care visits during treatment from 22.3% to 12.3%. This challenges a zero-benefit reading, but randomization tested additional visits among AI-selected patients, not AI selection against selection without AI.
- Publisher
- Journal of Clinical Oncology, indexed by PubMed
- Kind of source
- Peer-reviewed paper
- Describes events of
- Public from
Context only: Lund University,
fewer interval cancers
Why it matters here: MASAI supplies comparative evidence of better screening outcomes before the post. The reported interval-cancer counts do not themselves establish a reduction in deaths or suffering, and this release does not establish superiority statistically.
- Publisher
- Lund University
- Kind of source
- Independent measurement
- Describes events of
- Public from
Context only: Centers for Disease Control and Prevention,
619,871 people died of cancer.
Why it matters here: These retrospective data concern a period before the post and document the continuing burden. They neither measure AI exposure nor show what mortality would have been without AI.
- Where
- Overview, Fact
- Publisher
- Centers for Disease Control and Prevention
- Kind of source
- Official statistics
- Describes events of
- Public from
What would settle it
Comparative evidence measuring material mortality or suffering benefits attributable to AI through 2026-03-16, together with sufficiently broad coverage to assess the claimed absence of such benefits.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, F Cancer
AI is supposed to change all that. But so far it hasn’t.
- Where
- paragraph 2
- Published
- 2026-03-16
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-f-cancer-1e2f0f-a1: Current. Outcome: Unresolved. Evidence as of 2026-10-28. Public date not recorded.
Recorded events
Assessment publishedBy this site
First assessment published after both reviews and the final-text check passed.