A CERN for AI and the Global Governance of AI
We recorded 13 claims from this post. 4 were checked against evidence. 9 can’t be checked. 3 more are held. Read the post at its source.
Claims from this post (13)
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
Forecast
“The world can likely benefit enormously from AI, in ways that extend far beyond what we have already seen in recent months with productivity, but anyone with eyes can see that there are risks, too.”
Our reading: The world can likely benefit enormously from AI, in ways extending far beyond the productivity gains seen in recent months, though there are clearly risks too.
No horizon
Qualifier: can (potential benefit, not a prediction that it will happen)
Assessment: Can’t be checkedSaid
Observation
“What I noted in 2017 (at the time focusing on AI for medicine rather on AI safety per se) is that some problems in AI might be too complex for individual labs, and not of sufficient financial interest to the large AI companies like Google and Facebook (now Meta), ehich are after all, driven by profits rather than humanity. That’s still true.”
Our reading: Some AI problems may be too complex for individual labs and of insufficient financial interest to profit-driven large AI companies, which still holds true.
No horizon
Assessment: Can’t be checkedSaid
Observation
“And while every large tech company is making some effort around AI safety at this point, the collective sum of those efforts hasn’t yielded all that much.”
Our reading: The combined AI safety efforts of large tech companies have not yielded much.
No horizon
Assessment: SupportedSaid
Observation
“We have no way of guaranteeing that large language models will reliably stick to facts, a point so thoroughly established by now I don’t think it needs a reference.”
Our reading: There is no way to guarantee that large language models reliably stick to facts.
No horizon Evidence as of
Limitation: Support rests on a contemporary research synthesis, not an exhaustive proof of nonexistence.
Assessment: SupportedSaid
Observation
“We have no way of guaranteeing that large language models won’t be jailbroken into saying unethical things. To the contrary, a just-published technique shows that current LLMs can be jailbroken quickly, and automatically.”
Our reading: There is no way to guarantee LLMs won't be jailbroken, and current LLMs can be jailbroken quickly and automatically.
No horizon Evidence as of
Limitation: The guarantee clause depends on the stated general-immunity interpretation.
Assessment: UnresolvedSaid
Observation
“Current AI has no way to reliably anticipate consequences of actions that it might recommend, Asimov’s First Law of Robotics called for robots to do no harm, but current AI has no reliable way to anticipate harm.”
Our reading: Current AI cannot reliably anticipate the consequences or harms of actions it recommends.
No horizon Evidence as of
Limitation: The scope of current AI and the meaning of reliably are unspecified.
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“Anthropic’s “Constitutional AI” idea seems to me to be clever and the best of the bunch, but still inadequate, riddled with the usual problems that LLMs face (reliability, excessive dependence on similarity to what is in the training set, hallucinations, unreliable reasoning, etc).”
Our reading: It seems to Marcus that Anthropic's Constitutional AI idea is clever and the best of the current AI safety ideas, but still inadequate, riddled with the usual LLM problems such as unreliability, excessive dependence on similarity to training data, hallucinations and unreliable reasoning.
No horizon
Qualifier: seems to me
Assessment: SupportedSaid
Causal claim
“you don’t have to be superintelligent to create serious problems.The real issue is control.”
Our reading: AI need not be superintelligent to create serious problems; the real issue is control.
No horizon Evidence as of
Limitation: The example concerns automated driving, not an LLM.
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“We need to stop worrying (just) about Skynet and robots taking over the world, and think a lot more about what criminals, including terrorists, might do with LLMs, and what, if anything, we might do to stop them.”
Our reading: AI risk discussion should focus much more on misuse of LLMs by criminals and terrorists, not just on AI takeover scenarios.
No horizon
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“We need to consider both current AI risk and AGI risk. Period.”
Our reading: We need to consider both current AI risk and AGI risk.
No horizon
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“Don’t listen to them; enormous money is at stake.”
Our reading: People should not listen to those who deny there is any AI risk, such as Marc Andreessen, because enormous money is at stake.
No horizon
Assessment: Can’t be checkedSaid
Forecast
“There are over 60 elections around the world in 2024; the chance that at least one will be swayed by deep fakes is high.”
Our reading: It is likely that at least one of the 60+ elections worldwide in 2024 will be swayed by deepfakes.
No horizon
Assessment: Can’t be checkedSaid
Recommendation or value judgment
“Count me among those who continue to favor a CERN for AI safety and Global Safety, so long as we balance the portfolio of risks we are trying to address.”
Our reading: A CERN for AI safety and global AI safety governance should be created, provided they address a balanced range of risks.
No horizon
Qualifier: balance the portfolio of risks addressed
Held from this post (3)
These records did not pass every check, so nothing about them is published. Held is not a finding.
- The two reviews read the claim differently
- The record does not meet the current rules. Our reading, not assessed: We lack substantial controls over current AI, especially as it gets connected to real-world services and APIs.
- The two reviews read the claim differently