Reports of LLMs mastering math have been greatly exaggerated
We recorded 6 claims from this post. 6 can’t be checked. 1 more is held. Read the post at its source.
Claims from this post (6)
In the order they appear in the post. Each quotation is the speaker’s words in this post. A label such as “Supported” or “Can’t be checked” is the claim’s current assessment; its page has the evidence and the reasons.
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“The refusal of these kinds of AI to admit ignorance or incapacity and their obstinate preference for generating incorrect but plausible-looking answers instead are one of their most dangerous characteristics.”
Our reading: The refusal of these kinds of AI (LLMs) to admit ignorance or incapacity, and their obstinate preference for generating incorrect but plausible-looking answers instead, are among their most dangerous characteristics.
No horizon
Scope: these kinds of AI (LLMs like those tested on the USAMO)
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“Getting AIs to answer “I don’t know” is one of the most important unsolved challenges facing the field.”
Our reading: Getting AIs to answer 'I don't know' is one of the most important unsolved challenges facing the field.
No horizon
Assessment: Can’t be checkedSaid
Conditional forecast
“If this kind of technology becomes commonly used to answer difficult questions before the problem of generating invalid answers is fixed, we will be in serious trouble.”
Our reading: If LLMs become commonly used for difficult questions before their invalid-answer problem is fixed, society will be in serious trouble.
No horizon
Only if: LLMs become commonly used to answer difficult questions
Assessment: Can’t be checkedSaid
Forecast
“It is by no means certain that additional user prompts will add much,.”
Our reading: Additional user prompting may not substantially improve reasoning LLMs' performance on such problems.
No horizon
Qualifier: models already designed for complex reasoning
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“Moreover, however you slice it, this need of the AIs for special cajoling before they will think carefully about questions that are presented to them is a serious weakness, and the demand for special prompting tilts the scale.”
Our reading: However you slice it, the AIs' need for special cajoling before they will think carefully about questions presented to them is a serious weakness, and the demand for special prompting tilts the scale.
No horizon
Qualifier: however you slice it
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“The really important challenge is not to get the AIs to solve more USAMO problems; it is to get them to say “I give up” when they can’t.”
Our reading: The really important challenge is not getting AIs to solve more USAMO problems but getting them to say 'I give up' when they can't solve them.
No horizon
Held from this post (1)
These records did not pass every check, so nothing about them is published. Held is not a finding.
- The criteria are ambiguous. Our reading, not assessed: The authors have yet to see any evidence that any kind of prompt helps get AIs to say 'I give up' when they cannot solve a problem.