Prof in entrepreneurship & GenAI @Haskayneschool. Maker of https://teapaw.org/t.co/1BG4KFm3Jw, https://teapaw.org/t.co/z0ozbKNiJW, https://teapaw.org/t.co/IppyXplTi5, & https://teapaw.org/t.co/eG20WRZ5dH
Calgary, Alberta, Canada
Joined May 2008
- Tweets10.9K
- Following6.3K
- Followers2K
- Likes64.9K
Mohammad Keyhani retweeted
this is very clearly an interface and interaction design problem, not something inherent to agents. the system is designed in a way that encourages maximal decision deference to the machine, under the guise of machine intelligence. it is also designed in a way that encourages passive, disengaged interaction with the system.
anyone even remotely familiar with decision architecture and the decision support v automation distinction can instantly recognize that the way information is structured and presented, and the role the user is placed in within the user-machine interaction dynamic is one where the design intent is for the system to make the decision (as many as possible!), and the intended role of the user is to approve it -- ideally approving as few decisions as possible. this drives engagement in a facebook-y social media way, but leads to vastly worse outcomes.
the results are evident.
also -- while (obviously) I'm a big proponent of thinking deeply about how we rearchitect organizations and thoughtfully design work, I do think this particular issue is a design problem *first*, and an organizational architecture problem *only* insofar as organizations are forced to behave defensively against adversarial product design as long as better alternatives are not available.
Mohammad Keyhani retweeted
“We installed Khomeini in 1979. “
U.S. Ambassador to Türkiye Tom Barrack on Iran:
It's very difficult for the average Iranian to survive. Once they get out of a chokehold of that, over time they can change their own regime.
We've changed it three times, right? We installed the Shah. We installed Khomeini in 1979. We took out the Shah.
Source: @MarioNawfal
Looks like a major breakthrough in understanding LLM brains!
interpretability might be the most important scientific problem in the world right now, as I don't think we can solve alignment if we don't know what we're building. I think I have good news about it. I co-authored early work in mechanistic interpretability at OpenAI and came away pessimistic, because neural networks appeared to fold into arbitrary tangled knots. years later, I started openai's first large-scale automated interpretability project, finding about 2% of neurons legible with the primitive tools we had
with modern agentic tools and a much closer look, the picture looks totally different. the vast majority in models up to trillions of parameters are legible, and their circuits beautiful. many famous examples of tangled neurons turned out to be misunderstood. The roadblocks I expected mostly aren't there. I now think ambitious interpretability will be solved by the time we need it most, and possibly quite soon, as automated scientists keep improving. it may be a bitter lesson for interpretability: the simple thing, at absurd scale, works
already-clean.goodfire.com/
Mohammad Keyhani retweeted
There was a moment maybe two years ago when I realized that ChatGPT could have few-shot my phd dissertation (at least huge chunks of it). I did mourn for a few days, I worked my ass off for that thing.
Over time though I became sad for a different reason, that if I had ChatGPT during my PhD I could have tackle a far bigger, more important, more technically difficult dissertation question. Sometimes I think about the dissertation I could have written if I had Codex, something that an AI couldn't few-shot and I couldn't pull off without AI assistance.
Mohammad Keyhani retweeted
I've been prototyping an automated peer review project. With current models you get human level peer review for about $0.10 of tokens. For $5 you can rederive every result, check every statistical inference, and audit every paper cited in both directions.
There are about 100m papers in the scientific corpus. Most are not important. Many are sloppy. For structural reasons it's essentially impossible to change published work, even to correct a typo.
For $500m today you can thoroughly audit every paper ever written. Every p-hacked result, every plagiarist, every fabrication, every citation club, everything.
I don't have $500m so instead I build a tree, focusing on the 1% of papers that carry 99% of the signal, doing a cheap first pass looking for obvious problems. I can find 90% of the nonsense for a few hundred dollars of tokens.
Any guesses where the biggest surprises will be found?
Mohammad Keyhani retweeted
“On Tuesday night, I cried, and it woke my daughter. [OpenAI] annihilated my field of research. They wanted to prove they could massacre us.”
these are the words of Fields Medalist, Hugo Duminil-Copin, who was working on 16 problems OpenAI now claims it has solved.
and he says some mathematicians may now keep their research problems to themselves rather than share them publicly.
AI might actually make mathematics less open.
🇫🇷🤖🧑🔬 FLASH | « Ils ont annihilé mon domaine de recherche » : le mathématicien français Hugo Duminil-Copin, médaillé Fields, révèle avoir PLEURÉ après qu’OpenAI a publié des centaines de travaux prétendant résoudre des problèmes mathématiques encore ouverts, dont 16 sur lesquels il travaillait.
Il s’inquiète particulièrement de l’avenir des jeunes chercheurs et évoque un « démantèlement » de sa discipline. (Le Monde)
It has to be recognized that modern generative AI models are a product of our entire civilization and perhaps they should be taxed in a way that benefits our entire civilization.
Mohammad Keyhani retweeted
Prediction: the value of all verifiable tasks in the long run is 0. There is no profit there.
Anything where the success is defined in a finite and non-ambiguous way, it’s over. Value is 0.
The profits are all in taste/judgement related tasks where some like it and some hate it. Vibes only. No rationality.
Labs are hillclimbing verifiable tasks now and it might sound impressive that OpenAI found those math proofs but soon others will follow and players like DeepSeek will drive its cost to electricity.
Ambiguity is the life line and where the money is.
- Anything that’s irrational, vibe driven and controversial with no consensus definition of good is worth something.
- If the society all agree on what good looks like for a given task with no ambiguity, it’s worth 0.
- caveat: coming up with verifiable task to do is worth a lot because the ability to come up is not a verifiable task by itself
Mohammad Keyhani retweeted
From the blog:
'AI chips shipped through 2027 could run tens to hundreds of millions of concurrent frontier-model agents. Running nonstop, these agents would supply as many weekly working hours as about 140–720 million full-time employees.
More efficient models could potentially support billions of agents on the same hardware. Applying DeepSeek V4 Pro serving benchmarks to the projected hardware supply yields approximately 1.9 billion concurrent agents supplying as many weekly working hours as 8 billion people each working 40 hours.
We estimate AI infrastructure could soon support hundreds of millions or even billions of AI agents. At the high end, that’s enough to rival the working hours of the global human workforce.
The exact number of agents depends on the efficiency of the models they run on and whether soaring demand for AI continues. We analyzed different scenarios 🧵
ALT Chart: potential concurrent agents from 2025–27 memory shipments by model, from 30–60 million (Claude Fable 5) and 97–195 million (GPT-5.6 Sol) to 240 million (Kimi K3), 571 million (GLM-5.2) and 1.9 billion (DeepSeek V4 Pro).
Mohammad Keyhani retweeted
It can be hard to “feel the AGI” until you see an AI surpass you in a domain you care deeply about. This week, many mathematicians and physicists at @OpenAI had their Lee Sedol moment seeing this model solve, in minutes, open problems they’d struggled with for years.
This model represents a step-function improvement on many benchmarks, and its training is ongoing.
Our internal model group arrived at the Navier–Stokes solution in 88 hours, using around 10,000 coordinating AI agents.
Throughout the effort, we maintained the strict safeguards—including monitoring and isolation—that we apply to all our frontier evaluations.
Mohammad Keyhani retweeted
There’s a joke that goes something like “animals are a technology invented by plants to move seeds around” and this is the sense in which I think AI is a technology invented by humans to perform cognitive tasks.
Mohammad Keyhani retweeted
Can foreign attacks strengthen the regime they seek to weaken? Presenting empirical evidence that the 2026 US–Israeli strikes reduced anti-regime protest and boosted pro-regime mobilization, from @ProfSultanEcon, @lwantchekon, and Yaroslav Prokhorskoy nber.org/papers/w35734
Mohammad Keyhani retweeted
"Interviewees describe hacking little girls’ phones to see when their fathers came home and then bombing their houses, listening while the people on the line screamed and pleaded for help, while mothers cried over their burning daughters."
Please read this.
Yuval Abraham and Rachel Szor’s ‘Naza’ is one of the most important films of our time. 24 different Israeli intelligence officers testify in this unflinching exposé of systematic Israeli war crimes in Gaza. vulture.com/article/review-n…
TBH the UI for tasks with all projects being shown in the sidebar is much worse than before. Navigating between main and tasks is terrible now, and finding suggested tasks is way harder. Finding cancelled tasks has always been too hard because it's so easy to accidentally cancel.
Are scholars who do mathematical research in social science eligible? Really sad to see the social sciences and humanities so callously discarded by the AI labs.
Mohammad Keyhani retweeted
This might not make me popular but I remain continuously impressed by Carney: to me, one of the few (if not the only) genuine statesman in the West.
It genuinely takes incredible balls and statesmanship - especially for Canada, arguably the western country that's most dependent on the US - to walk away from the table and match Trump's tariffs China-style.
Europe, with far less dependence on the US and considerably more resources than Canada, just pathetically caved and signed on Trump's humiliating terms when in the same situation.
