“The thing about reinforcement learning is that it is notoriously finicky; change the circumstances sligthly, and it may no longer work.”
Our reading:
Reinforcement learning is notoriously finicky and may stop working when circumstances change slightly.
Assessment: Can’t be checked
What happened
Not graded on evidence
Outcome: Can’t be checked
An opinion, a recommendation, or too vague to test. Listed, not scored.
Evidence up to . Checked .
Why
The operative wording is 'may no longer work.' The supplied policy excludes possibility language and does not permit the plain-reading exception for hedged claims. Treating this as an assertion that documented reinforcement-learning vulnerabilities exist would repeat the interpretation the reviewers rejected. The accompanying phrase 'notoriously finicky' supplies no independent scoring standard.
Caveats
- This decision does not assess the research literature on reinforcement-learning robustness.
- The surrounding Atari example and conjecture about Microsoft's implementation are not separately graded.
- Outcome-known status is false because no scoreable outcome is defined.
Evidence (0 items)
Claims like this aren’t scored, so there is no evidence to show.
How we judged it
- Kind of claim
- Observation
- Deadline
- None given. We don’t add one.
Conditions
- when circumstances change slightly (their words)
Where we had to interpret
- Marcus describes a possible failure under a slight change in circumstances, not a necessary or quantified failure rate.
- 'Notoriously finicky' is a qualitative characterization without a stated performance standard.
- Under the supplied policy, the possibility clause cannot be converted into a scoreable empirical vulnerability assertion.
- The surrounding conjecture about Microsoft's system is not treated as an assertion that reinforcement learning caused Bing's behavior.
We don’t grade claims like this. We only check the quote and who said it.
We found 14 claims in this post. See them all.
Full record
Source and dates
Marcus on AI, Why *is* Bing so reckless?, paragraph 32.
- 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 2023-02-20.
Where it was said (1)
Said once in the posts we read.
: Marcus on AI, Why *is* Bing so reckless?
The thing about reinforcement learning is that it is notoriously finicky; change the circumstances sligthly, and it may no longer work.
- Where
- paragraph 32
- Published
- 2023-02-20
- We read it
- , live page
- Access
- full text, public
History
Versions
- Current: Can’t be checked. 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.