OG GenAI Skeptic; spoke at US Senate. Warned about hallucinations in 2001. Advocating world models & neurosymbolic AI ever since. Author, Marcus on AI & 6 books
Joined December 2010
- Tweets64.3K
- Following7.2K
- Followers255K
- Likes95.7K
Pinned Tweet
Three thoughts on what really matters:
1. Fuck cancer
2. Friends are irreplaceable
3. The new "Marcus test" for AI is when AI makes a significant dent on cancer
May that happen sooner, much sooner, rather than later.
In memory of my childhood friend Paul.
I got to meet the Pope! Report from the Vatican, with highlights from @Pontifex’s terrific speech yesterday and some thoughts of my own:
garymarcus.substack.com/p/re…
Looking like this tweet still has some life to it…
9 reasons that OpenAI could someday be seen as the WeWork of AI:
👉 Lots of competitors are catching up.
👉 OpenAI has been forced to cut prices.
👉 A number of key employees have departed.
👉 Meta’s newest model is likely to be on par with GPT-4 and free.
👉 After 5 months, Sora still hasn’t shipped
👉 GPT-5 is long delayed, likely at least until 2025.
👉 Even if GPT-5 ships, it will likely be quite costly to run.
👉 Core problems of reliability & factuality aren’t likely to be solved.
👉 They have never turned a profit; and they might never.
Gary Marcus retweeted
Less than a month ago, OpenAI was looking to raise “at least” $30 billion at a valuation of $1.4 trillion.
Today it raised $11 billion at $1.25 trillion valuation and faced “tepid” interest.
OpenAI has raised less than half of the $30 billion it had initially targeted for its latest funding round on.wsj.com/4zqPW7W
elon who is doing his best to make America a “permanent one-party state” complains, with absolutely zero self-awareness, that Democrats want to make America a “permanent one-party state”.
My biggest concern is that if there is a clean sweep by Democrats of the House, Senate and Presidency in 2028, they will:
1. Make DC and Puerto Rico states to gain a strong majority in the Senate.
2. Open the borders and legalize everyone living in America illegally to cement a permanent majority.
3. Increase the size of the Supreme Court to gain an immediate majority.
This would result in America becoming a permanent one-party state and the end of democracy in any meaningful sense of the word.
All that is needed to ensure this doesn’t happen is 51 Republican senators in 2029. That is why you should vote for Republican senators this election. Other reasons are irrelevant compared to this one.
Gary Marcus retweeted
In my latest Expert Voices column for @ScienceMagazine, I weigh in on this summer's - and this week's - AI for math news:
AI has solved many math problems, but it has not solved math | Science science.org/doi/10.1126/scie…
Gary Marcus retweeted
Was the OAI Navier-Stokes solution a result of specification gaming?
Claim: OAI did satisfy the letter of statements with a mathematically valid construction, but it relies on an artificial force so extreme that it tells us nothing about whether the equations remain well-behaved under the conditions that occur in our universe. If so, then you might be able to say that the solution does not satisfy the intended scientific question.
This quoted post is unavailable.
Ant’s IPO is based on fantasy projections:
.@econJaredB and I wrote about Anthropic's IPO. To justify a $2T valuation, they would need to have revenue in 2031> $500B, which is 14x greater than Google's revenues 5 years after their IPO and more than twice Meta's current revenue.
There is a very real chance that Anthropic will at some point fail to be able to pay its bills.
Blackstone just figured that out.
“.. a reversal from its previous guidance that would complicate the AI lab’s ability to finance its massive spending.”
@FT $ANT
ft.com/content/59383c7f-8d82…
Gary Marcus retweeted
In this new paper, we make a strong point:
Human imitation is a misplaced goal for AI.
Too much of AI development is based on human imitation, as exemplified by the Turing test: the more a machine resembles us, the more successful it is.
But there are two major problems.
The first is conceptual. Imitating human outputs does not tell us much about the capacities behind them:
The same answer can emerge from different processes. A plausible response does not tell us whether a system can reason reliably, generalize to unfamiliar situations, or recognize when it should say “I don’t know”.
The second is ethical. Why should we want AI that imitates humans?
Humans are biased. Our attention is limited. Our judgment is fallible. Reproducing these limitations is hardly an ambitious vision for AI.
Even worse:
A system that impersonates a person can exploit human trust.
And an industry focused on replicating humans can fall in the “Turing’s trap”: favour human replacement over augmentation.
Instead, we argue that AI should aim to improve human welfare.
To this end, we propose three coordinated changes:
For developers: Build systems that strengthen human capabilities and independent judgment.
For funders: Require projects to assess benefits, costs, and who receives or bears them.
For institutions: Base procurement and renewal on demonstrated benefits.
The goal should be to help people know more, suffer less, and exercise greater control over their lives.
*
Full paper in the first reply.
With @GaryMarcus & @Walter4C
Gary Marcus retweeted
The future of #ArtificialIntelligence is not pre-determined. As a technology developed by human beings, its future depends on the choices we make. We therefore have the responsibility to guide, change and, when necessary, reject its applications. A reliable system that serves the human person is of greater value than one that is exceptionally powerful but whose consequences cannot be adequately foreseen or controlled.
Gary Marcus retweeted
I actually believe this data to be consistent with the broken nature of the IPO process. Instead of seeking a true market price they are trying to reward their clients w big one day pops. Those charts show retail getting taken to the cleaners.
Simply magical. Especially in light of @Pontifex’s pitch-perfect moral leadership on AI.
Gary Marcus retweeted
A useful razor for evaluating the current moment, especially in regards to tech, is considering how stupid it's going to look in 5, 10, or 25 years. I can't help but think this will all seem like an age of hucksters and manipulators. And yeah, moreso than usual.
why does anyone even listen to him anymore?
"I thought the alignment problem would turn out to be harder."
Pretty wild thing for Sam Altman to say in light of Hugging Face and findings that Astra would totally hack and social engineer when accidentally given access to a (secretly simulated internet) from UK AISI.
He caveats by saying it will probably get harder, but his answer suggests they have figured out how to align today's models, when that's clearly not true. Curious how many OpenAI researchers have seen this and what they think.
