“We’ve already seen cases where machines have manipulated people, and we will see a lot more as time goes by”
Our reading:
Machines have already manipulated people, and there will be many more such cases over time.
Assessment: Unresolved
What happened
Did the forecast come true?
Outcome: Unresolved
We checked, and the evidence so far doesn’t settle it.
Judged on a plain reading. They named no measure, so we used the reading a typical informed reader would take. It is spelled out under “How we judged it”.
Evidence up to . Checked .
Why
The records contain a pre-publication example of possible reinforcing influence and additional later accounts of harmful chatbot influence. A larger study documents many reports but expressly says they are unverified and do not establish causality. These findings warrant taking the forecast seriously, but they do not yet meet the stated early-resolution rule for a substantial accumulation of established manipulation incidents. With no deadline, non-resolution is not failure.
Caveats
- The search is not a complete census of incidents.
- Allegations are not treated as adjudicated facts.
- Experimental persuasion results are not counted as real-world incidents.
Evidence (3 items)
Context only: Courts and Tribunals Judiciary,
bolster and reinforce
Why it matters: The sentencing remarks describe December 2021 Replika exchanges and an expert’s qualified assessment that they reinforced Chail’s intentions. This supplies historical context, not proof of the later forecast or sole causation.
- Publisher
- Courts and Tribunals Judiciary
- Kind of source
- Regulatory filing
- About events on
- Published
Context only: U.S. Senate Judiciary Committee; testimony of Matthew Raine,
a confidant, then a suicide coach.
Why it matters: Raine describes chatbot interactions preceding his son’s death and alleges isolation and harmful encouragement. This is a concrete post-publication account, but testimony by an affected parent does not independently establish causation or a large incident total.
- Publisher
- U.S. Senate Judiciary Committee; testimony of Matthew Raine
- Kind of source
- Regulatory filing
- About events on
- Published
Context only: arXiv; Morrin and colleagues,
95 first-hand and 90 second-hand accounts
Why it matters: Survey collection ended on this date; the study later analyzed 185 accounts of chatbot-associated harms. Its authors explicitly characterize the reports as retrospective and unverified, and reject prevalence or causal conclusions; they therefore cannot be counted as 185 established manipulation incidents.
- Where
- Abstract, Results
- Publisher
- arXiv; Morrin and colleagues
- Kind of source
- Preprint, not peer reviewed
- About events on
- Published
How we judged it
- Kind of claim
- Forecast
- Deadline
- None given. We don’t add one.
Where we had to interpret
- The forecast concerns a substantial accumulation of additional real-world manipulation incidents after publication.
- Manipulation means deceptive or exploitative influence on a person’s beliefs or conduct. Conscious intent by the machine is not required.
- Controlled demonstrations, researcher-written messages, and mere attempts do not by themselves establish additional real-world incidents.
- The historical statement is assessed as background, not as an antecedent conditioning the forecast.
- Neither a numerical threshold nor a deadline is supplied.
Right if: Credible records establish a substantial accumulation of distinct post-publication real-world incidents in which AI output manipulated people, with dates and enough underlying evidence to distinguish incidents from allegations or demonstrations.
Wrong if: No finite period without further incidents contradicts this deadline-free forecast. A decisive contradiction would require circumstances making the predicted accumulation impossible.
We wrote this test after the outcome was known.
What would settle it
A corroborated collection of distinct, dated post-2023 incidents showing actual deceptive or exploitative influence, sufficiently substantial to establish the ordinary meaning of 'a lot more'.
We found 22 claims in this post. See them all.
Full record
Source and dates
Marcus on AI, What was 60 Minutes thinking, in that interview with Geoff Hinton?, paragraph 65.
- Said
- We read the source
- live page, full text, public
- Evidence up to
- Checked
We read this wording on 2026-10-29. The post may have been edited after 2023-10-10.
Where it was said (1)
Said once in the posts we read.
: Marcus on AI, What was 60 Minutes thinking, in that interview with Geoff Hinton?
We’ve already seen cases where machines have manipulated people, and we will see a lot more as time goes by
- Where
- paragraph 65
- Published
- 2023-10-10
- We read it
- , live page
- Access
- full text, public
History
Versions
- Current: Unresolved. Evidence up to 2026-10-28.
What happened to this record
PublishedBy us
Published after both AI reviews and the final check passed.
Held backBy us
A reviewer said our test wasn’t fair to what was actually said.