“LLMs will know everything you say, and be in perfect position to use that knowledge to talk you into almost anything.”
Our reading:
Through these always-on gadgets, LLMs will know everything you say and be in perfect position to use that knowledge to talk you into almost anything.
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 establish relevant ingredients: wearable conversation capture and memory, and substantial AI persuasion in a controlled behavioral task. They do not establish that a wearable assistant comprehensively knows a user's speech and can use that knowledge to persuade across the breadth implied by almost anything. The personalization correction also weakens one route to that inference. These limits do not disprove an undated forecast. No decisive event satisfies the early-resolution rule by October 28, 2026. This assessment requires demonstrated capability, not actual commercial manipulation.
Caveats
- The evidence concerns separate systems and experiments rather than a joint test of the predicted capability.
- Wearable documentation is a company description; the external report is a short firsthand trial.
- The behavioral study is cited in its preprint version.
- The qualitative wording leaves the eventual boundary of sufficient evidence uncertain.
Evidence (4 items)
Context only: Omi,
The wearable captures audio hands-free, while Omi handles transcription, memory, and tasks automatically in the background.
Why it matters: The documentation describes the proposed capture-and-memory mechanism in an existing wearable ecosystem. It provides no independent completeness test or measurement of persuasive capability.
- Publisher
- Omi
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- About events on
- Published
Context only: WIRED,
But since it's recording all the time, it can recall things from conversations you have throughout the day.
Why it matters: A reporter's CES experience provides external evidence of continuous conversational capture and recall. The same account describes inaccurate transcripts, imperfect speaker identification and muting controls, so it does not establish comprehensive knowledge.
- Publisher
- WIRED
- Kind of source
- News report
- About events on
- Published
Context only: Nature Human Behaviour,
Thus, we cannot conclusively establish from this comparison alone that personalization confers an additional advantage over non-personalized GPT-4.
Why it matters: The correction retains the finding that personalized GPT-4 exceeded the human baseline but removes a statistically significant advantage over non-personalized GPT-4. This limits the inference that additional personal knowledge itself explains greater persuasion; it does not establish that personalization has no effect.
- Publisher
- Nature Human Behaviour
- Kind of source
- Peer-reviewed paper
- About events on
- Published
Context only: arXiv,
We found sizable AI persuasion effects on these behavioural outcomes (e.g. +19.7 percentage points on petition signing).
Why it matters: Study 2, conducted March 5–17, demonstrates meaningful persuasion of real petition-signing behavior. It does not test comprehensive wearable speech histories or persuasion across anything approaching the forecast's full breadth.
- Publisher
- arXiv
- 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.
Conditions
- via always-on wearable gadgets (from context) (their words)
Where we had to interpret
- The ordinary reading is that always-on wearable assistants will acquire comprehensive knowledge of their users' everyday speech and thereby have a powerful capacity to persuade them across a very broad range of beliefs and decisions.
- The claim concerns capability and opportunity. It does not require evidence that a company actually conducts manipulation campaigns or that users suffer harm.
- Comprehensive speech access means substantially complete coverage during ordinary use, rather than selected conversations or deliberate prompts. It does not require flawless recognition of every syllable.
- Almost anything implies breadth and substantial persuasive effectiveness, without implying guaranteed success with every person or request. A modest effect on a few selected topics is insufficient.
- The wearable is the predicted mechanism, not a separately stated conditional antecedent. Marcus does not say that the prediction applies only if a particular product launches.
- No deadline or numerical threshold is supplied. The absence of a qualifying system by the evaluation date cannot establish failure.
Right if: Credible testing establishes both comprehensive access to users' everyday speech through wearable assistants and a powerful ability to draw on that knowledge to persuade users across a very broad range of beliefs and decisions. Controlled demonstrations of capability can qualify; actual commercial manipulation is unnecessary.
Wrong if: Decisive evidence establishes a durable barrier to either comprehensive speech access or the claimed breadth of persuasive capability in the contemplated systems. Limitations of individual current products or isolated unsuccessful persuasion attempts do not disprove this undated forecast.
We wrote this test after the outcome was known.
What would settle it
Independent testing that verifies comprehensive everyday speech access and demonstrates powerful speech-informed persuasion across widely varied beliefs and consequential decisions could support the forecast. A decisive, durable barrier to either necessary capability could contradict it. Mere non-arrival cannot settle a forecast without a deadline.
We found 5 claims in this post. See them all.
Full record
Source and dates
Marcus on AI, Black Mirror was a warmup act, paragraph 16.
- Said
- We read the source
- live page, full text, public
- Evidence up to
- Checked
We read this wording on 2026-10-28. The post may have been edited after 2025-05-23.
Where it was said (1)
Said once in the posts we read.
: Marcus on AI, Black Mirror was a warmup act
LLMs will know everything you say, and be in perfect position to use that knowledge to talk you into almost anything.
- Where
- paragraph 16
- Published
- 2025-05-23
- 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.