The Sparks of AGI? Or the End of Science?
We recorded 7 claims from this post. 3 were checked against evidence. 4 can’t be checked. 4 more are held. Read the post at its source.
Claims from this post (7)
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
Conditional forecast
“That claim would never survive serious scientific peer review.”
Our reading: Microsoft's claim that GPT-4 is an early (yet still incomplete) version of an AGI system would never survive serious scientific peer review.
No horizon
Only if: if the claim were subjected to serious scientific peer review
Assessment: SupportedSaid
Observation
“The problem of hallucinations is not solved; reliability is not solved; planning on complex tasks is (as the authors themselves acknowledge) not solved.”
Our reading: GPT-4 has not solved hallucinations, reliability, or planning on complex tasks.
No horizon Evidence as of
Limitation: The planning example concerns the early research model examined in Sparks.
Assessment: Can’t be checkedSaid
Observation
“As I have said before, I don’t really think GPT-4 has much to do with AGI.”
Our reading: Marcus does not really think GPT-4 has much to do with AGI, as he has said before.
No horizon
Qualifier: I don't really think
Assessment: UnresolvedSaid
Causal claim
“The paper’s core claim— “GPT-4 attains a form of general intelligence [as] demonstrated by its core mental capabilities (such as reasoning, creativity, and deduction)”—literally cannot be tested with serious scrutiny, because the scientific community has no access to the training data.”
Our reading: The Sparks of AGI paper's claim that GPT-4 attains a form of general intelligence cannot be seriously tested because scientists have no access to its training data.
No horizon Evidence as of
Limitation: This is not support for the claim that all serious testing requires training-data access.
Assessment: MixedSaid
Observation
“Microsoft and OpenAI are producing and deploying products with potentially enormous risks at mass scale, and making extravagant, unreviewed and unreviewable claims about them, without sketching serious solutions to any of the potential problems they themselves have identified.”
Our reading: Microsoft and OpenAI are producing and deploying products with potentially enormous risks at mass scale, making extravagant, unreviewed and unreviewable claims about them, without sketching serious solutions to any of the potential problems they themselves have identified.
No horizon Evidence as of
Limitation: The result treats substantive mitigations as satisfying sketching serious solutions.
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“This cannot and should not stand.”
Our reading: Microsoft and OpenAI's opaque deployment of risky AI systems should not be allowed to continue.
No horizon
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“We must demand transparency, and if we don’t get it, we must contemplate shutting these projects down.”
Our reading: The public must demand transparency from Microsoft and OpenAI, and if it is not provided, consider shutting their projects down.
No horizon
Qualifier: if transparency is not provided, contemplate shutting projects down
Held from this post (4)
These records did not pass every check, so nothing about them is published. Held is not a finding.
- The two reviews reached different assessments. Our reading, not assessed: The two giant OpenAI and Microsoft papers have been about a model about which absolutely nothing has been revealed, neither the architecture nor the training set.
- The two reviews read the claim differently
- The record does not meet the current rules. Our reading, not assessed: Microsoft and OpenAI are rolling out extraordinarily powerful yet unreliable systems with multiple disclosed risks and no clear measure either of their safety or of how to constrain them.
- The criteria are ambiguous. Our reading, not assessed: OpenAI's incorporation of user experiments into its training corpus destroys the scientific community's ability to test whether its models generalize to new cases.