“Our view is that they will not work well until major progress is made in common sense, which itself may require wholesale revisions in how AI is approached.”
Our reading:
AI agents will not work well until major progress is made in commonsense reasoning, which may require wholesale revisions to how AI is approached.
Assessment: Unresolved
What happened
Did the forecast come true?
Outcome: Unresolved
We checked, and the evidence so far doesn’t settle it.
Evidence up to . Checked .
Why
The claim says AI agents will not work well until major progress is made in common sense. It gives no deadline. Under the contract, the claim is contradicted early only if agents are shown to reliably complete realistic multi-step tasks and it is documented that no major commonsense advance occurred. It is supported early only if a recognised commonsense breakthrough is documented as what made agents work well. The cited evidence shows neither event. In the paper's September 2025 version, the best agent on TheAgentCompany completed 30% of tasks autonomously (e1). In July 2025, OpenAI said its ChatGPT agent was still in its early stages and could still make mistakes (e2). Neither item shows agents working reliably. Neither item addresses whether any commonsense progress happened. With no deadline, continued shortfalls do not settle the forecast. The hedged part, that progress 'may require wholesale revisions' in how AI is approached, cannot be tested on its own.
Caveats
- There is no agreed measure of 'working well' for agents and no agreed marker of 'major progress in common sense'.
- The evidence covers one benchmark and one vendor statement. It does not cover agent deployment more broadly.
- Neither evidence item looks at commonsense reasoning, so the record says nothing about the claim's condition.
Evidence (2 items)
Context only: arXiv (Xu et al., Carnegie Mellon University),
find that the most competitive agent can complete 30% of tasks autonomously.
Why it matters: By September 2025 the best agent still completed only 30% of the workplace tasks. This shows modest progress, but it says nothing about whether commonsense progress happened.
- Where
- Abstract (v3)
- Publisher
- arXiv (Xu et al., Carnegie Mellon University)
- Kind of source
- Preprint, not peer reviewed
- About events on
- Published
Context only: OpenAI,
ChatGPT agent is still in its early stages. It’s capable of taking on a range of complex tasks, but it can still make mistakes.
Why it matters: The vendor itself describes its mid-2025 agent as early-stage and error-prone. This is context on 'working well'.
- Publisher
- OpenAI
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- About events on
- Published
How we judged it
- Kind of claim
- Conditional forecast
- Deadline
- None given. We don’t add one.
Conditions
- until major progress is made in common sense (their words)
Where we had to interpret
- The claim links two things: agents working well, and major progress in common sense. To test it, you would need to establish both whether agents work well and whether major commonsense progress has occurred. Neither has an agreed measure.
- No deadline is given, and 'may require wholesale revisions' is hedged.
Right if: Agents keep failing on realistic tasks while there is no agreed major advance in machine common sense. Or agents come to work well only after a recognised commonsense breakthrough.
Wrong if: Agents are widely shown to work well on realistic tasks while experts agree no major commonsense progress has occurred.
We wrote this test after the outcome was known.
What would settle it
Independent benchmarks or deployment studies showing agents reliably completing a large majority of realistic multi-step tasks (for example, TheAgentCompany-style tasks). These would need to come with documented expert assessments either that no major commonsense advance occurred, or that a recognised commonsense breakthrough caused the improvement.
We found 16 claims in this post. See them all.
Full record
Source and dates
Marcus on AI, AI still lacks “common” sense, 70 years later, paragraph 30.
- 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 2025-01-05.
Where it was said (1)
Said once in the posts we read.
: Marcus on AI, AI still lacks “common” sense, 70 years later
Our view is that they will not work well until major progress is made in common sense, which itself may require wholesale revisions in how AI is approached.
- Where
- paragraph 30
- Published
- 2025-01-05
- 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
The final check failed: the write-up said more than the evidence showed.