Hard-forked! Casey Newton’s distorted portrait of Gary Marcus and AI skepticism
We recorded 14 claims from this post. 3 were checked against evidence. 11 can’t be checked. 14 more are held. Read the post at its source.
Claims from this post (14)
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: UnresolvedSaid
“Even if the AI acceleration we saw from 2019-2023 slows down for five years, as I have often speculated, it won’t slow down forever.”
What we tested: The AI acceleration seen from 2019-2023 will not slow down forever, even if it slows down for five years as Marcus has often speculated.
Assessment: Can’t be checkedSaid
“We ought to prepare – now – for AI to get a lot more dangerous than it already is.”
What we tested: Society should prepare now for AI becoming much more dangerous than it already is.
Assessment: Can’t be checkedSaid
“If there is even a 5% chance I am wrong about diminishing returns, we still ought to be moving immediately to prepare.”
What we tested: Even a 5% chance that LLM returns are not diminishing means society should move immediately to prepare for advanced AI.
Assessment: Can’t be checkedSaid
“We need, for example, a global treaty, and those tend to take a decade. Let’s get moving, people.”
What we tested: A global treaty is needed, as one example of preparation for more dangerous AI, and work on it should start now because such treaties tend to take a decade.
Assessment: Can’t be checkedSaid
“If AGI dropped tomorrow, given the almost total lack of regulation with real teeth, I seriously doubt we as a society would be ready for it.”
What we tested: If AGI arrived immediately, society would very likely not be ready for it, given the near-absence of regulation with real teeth.
Assessment: Can’t be checkedSaid
“As AI comes to have more power in society, AI skepticism becomes an increasingly important counterweight, and we should all want AI skepticism to be done right.”
What we tested: As AI gains power, AI skepticism becomes an increasingly important counterweight and should be done well.
Assessment: Can’t be checkedSaid
“AI needs substantive regulation.”
What we tested: AI needs substantive regulation.
Assessment: UnresolvedSaid
“The whole scaling debate (and the discussion around economics that Ed Zitron and I each often raise) is really about whether generative AI in particular can be made into AGI just using more data and compute; a system like AlphaFold 3 that combines classical AI with generative AI is on a different axis that has seen far less investment.”
What we tested: The scaling debate concerns whether generative AI alone can reach AGI with more data and compute, while hybrid systems like AlphaFold 3 have received far less investment.
Assessment: Can’t be checkedSaid
“Likewise, an AI with a known tendency to hallucinate things can be used to target weapon (this as may happen soon, given the recent announcement that OpenAI will collaborate with Anduril).”
What we tested: Hallucination-prone AI may soon be used to target weapons, given OpenAI's collaboration with Anduril.
Assessment: Can’t be checkedSaid
“I also always say we need hybrid systems that include LLMs as one tool among many, never that that LLMs will altogether disappear, even if they become a commodity and valuations fall.”
What we tested: AI needs hybrid systems that include LLMs as one tool among many.
Assessment: Can’t be checkedSaid
“LLM’s in particular are wildly overrated.”
What we tested: LLMs in particular are wildly overrated.
Assessment: Can’t be checkedSaid
“Many downsides are already heere, and potential misuse in the military seems imminent.”
What we tested: Military misuse of LLMs seems imminent, adding to downsides such as covert racism, deepfakes and discrimination that already exist.
Assessment: SupportedSaid
“LLMs are already dangerous, despite their limits.”
What we tested: LLMs are already dangerous despite their limitations.
What happened: The researchers demonstrated unfavorable job assignments and harsher hypothetical criminal judgments associated with African American English. This establishes a concrete discriminatory risk, while the experiment does not establish that employers or courts actually imposed these outcomes. (arXiv, )
Assessment: Can’t be checkedSaid
“Newton is of course right that we need to prepare for AGI, whenever it comes.”
What we tested: Society needs to prepare for AGI, whenever it arrives.
Held from this post (14)
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: LLMs still have trouble with causality, physical, psychological and mathematical reasoning, even if they now pass the specific published examples.
- The criteria are ambiguous. Our reading, not assessed: LLMs do not come close to delivering what was initially promised.
- The two reviews reached different assessments. Our reading, not assessed: Excessive enthusiasm for LLMs has drawn attention and resources away from other parts of AI.
- The record does not meet the current rules. Our reading, not assessed: Returns on LLMs are likely diminishing now or will very soon, though this is not certain.
- The record does not meet the current rules. Our reading, not assessed: None of the LLM weaknesses in causal, physical, psychological or mathematical reasoning has been robustly solved.
- The criteria are ambiguous. Our reading, not assessed: AGI will eventually arrive but not soon, and will require new ideas because the field's current hypothesis has basically been disproven.
- The two reviews read the claim differently
- The two reviews read the claim differently
- The record does not meet the current rules. Our reading, not assessed: An AI being stupid or overrated does not stop it being dangerous; such AI can be misused for propaganda and deepfakes.
- The two reviews reached different assessments. Our reading, not assessed: Recent statements from industry leaders like Andreessen, Nadella and Pichai confirm Marcus's earlier predictions about LLM limitations.
- The record does not meet the current rules. Our reading, not assessed: Many students, perhaps most of OpenAI's users, use LLMs daily to write term papers, and coders use them heavily too.
- The two reviews read the claim differently
- The two reviews read the claim differently
- The two reviews read the claim differently