Founder of the Natural Philosophy Institute (https://xitter.twitterapi.xyz/t.co/RsN3f31Sna) Betting against the 20th century. https://xitter.twitterapi.xyz/t.co/RZuOMXxvmV
Joined August 2014
- Tweets4.2K
- Following579
- Followers11.9K
- Likes3.9K
Pinned Tweet
The biggest undiscovered bet in the world is to bet against mathematicians.
I think I just figured out how to do it.
Propagandists saw the bat signal. AI coup underway.
And his should not be the only head to roll economist.com/leaders/2026/1…
5
387
Steve Patterson retweeted
I now assign some non-zero probability to the timeline where coding gets solved, solved swiftly, utterly, superexponentially, to the point that <30B models in Q2 2027 can zero-shot build operating systems – and then there's no "RSI" or singularity.
basically Chollet's take
One of the biggest misconceptions people have about intelligence is seeing it as some kind of unbounded scalar stat, like height. "Future AI will have 10,000 IQ", that sort of thing. Intelligence is a conversion ratio, with an optimality bound. Increasing intelligence is not so much like "making the tower taller", it's more like "making the ball rounder". At some point it's already pretty damn spherical and any improvement is marginal.
Now of course smart humans aren't quite at the optimal bound yet on an individual level, and machines will have many advantages besides intelligence -- mostly the removal of biological bottlenecks: greater processing speed, unlimited working memory, unlimited memory with perfect recall... but these are mostly things humans can also access through externalized cognitive tools.
Steve Patterson retweeted
In Neural Networks are Already Clean, @nickcammarata argued that neurons were far more legible than previously thought – and that this was great news for interpretability and safety.
Now, we’re releasing the first technical companion to Nick’s essay. It shares our findings from studying thousands of neurons using our novel automated interpretability pipeline, and outlines why we now think of neurons by their computational function, instead of whether or not they are polysemantic. Here’s what we found: 🧵
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/
Steve Patterson retweeted
Lightning Network capacity peaked at the beginning of 2026 and has dropped 45% in BTC terms year to date.
xitter.twitterapi.xyz/lnbits/status/21194288…
I remember reading this a decade ago and thinking it was correct and universally desirable.
Now I'm like "Absolutely correct and of the highest importance... for less than 10% of people."
The problem, as I see it, is not deference to authority. It's incompetent authorities.
2
847
Language is a part of physics and math. And neuroscience.
Language has order, structure, physical embeddings and physical consequences. It mechanically does stuff to your brain.
A lot of theorizing is brain-structure dependent. Taking the inclinations of one's thought--formed by one's individual brain structure--and generalizing into universals.
This seems to apply to everything from morality to geometry.
The way things are <-the way my brain is.
eg Utilitarianism might just be the philosophy of people with brain wiring X.
Reductive physicalists might be the philosophy of brains which fail to grasp abstractions.
Less about ideas and more about brain wiring.
1
2
347
Steve Patterson retweeted
🚨 We are receiving reports of hot wallets on BTCPay Servers being swept by unauthorized attackers.
If you're running an instance, please shut it down (./btcpay-down.sh) until we complete the investigation.
Yes. The priceless wabi-sabi of human speech and thought.
I'll be part of the human remnant fwiw. No brain chips, and no passing off AI writing as our own.
AI is an extraordinary aid to human reason, but it's not a replacement. Or at least, not for the remnant.
It turns out that using LLMs as an aid does make your presentations prettier and even your poems more highly rated, but it also turns out that these improvements come at the cost of “sameness.” Man, every gosh darn YouTube video sounds exactly the same! It’s the same pattern over and over. It’s an improvement in one sense, but so generic and frustrating. It has me longing for people who speak inarticulately because at least it’s coming from them. You can spot the LLM tells from a mile away, despite someone trying to hide them by changing em dashes to parentheses or using whatever other tricks to gasconade as original. The sweeping uniformity is there in the bloody bones. In the bloody structure. Everyone sounds the bloody same.
1
8
748
Steve Patterson retweeted
I actually think the lack of consensus about whether "real", randomness "exists" arises at least partly from a widespread misunderstanding of the mathematical definition itself.
For starters, the actually-correct definition appeared in 1966 (by Martin-Löf). That's 11 years after Einstein died, 33 years after Kolmogorov's measure-theoretic basis for probability theory (1933), and 40 years after Schrödinger/Born introduced wave mechanics and the probabilistic wave function (1927). During all that time, there WAS animated debate about the nature of "randomness" in physics, statistics, probability theory, and to some extent, math. But, how substantive could it have really been? They did not even have a clear picture of what the idea meant. And actually, since randomness is absolutely fundamental to all notions of probability, they ALSO did not even have good reasons to believe their work was well-posed at all.
But that's not even the end. It took another 9 years before the implications of this work really started to reach mainstream scientific consciousness. Knuth published an somewhat animated account of the struggle to reach the definition in TAOCP (1972), and then, 3 years later, Chaitin published a Scientific American article that sort of challenged empiricism. (See screenshot.)
So it took roughly 50 years of debate before scientists hadthe tools they needed to really have informed opinions about what they were arguing about. And even then, it's not obvious to me that this really happened. It seems to me that the fact that randomness is fundamentally UNCOMPUTABLE and in fact requires INFINITE information as INPUT, is very much germane to the question of whether whether it's physically plausible. Would we expect that the mechanism for producing quantum randomness accomplishes the same thing as the algorithm, but without (say) the infinite resource consumption? Presumably no. But in a survey I read just before posting the question ~60% of respondents appeared to believe it was a firm theoretical basis for quantum mechanics anyway!
I'm not saying there are 0 physicists who actually appreciate these subtleties, but I also do not sense it is widely appreciated as being important to the day-job of a physicist working on, say, the LHC. So for this reason, I'm not sure it will ever reach the level beyond mathematical curiosity. Oh well.
Steve Patterson retweeted
Update on this: Dad is checking all the calculations by hand (because he still doesn't fully trust the robot) and found a place where it should have used cos(x+a) instead of cos(x); the two are not the same since there's rotational variation
From 2023. I think it aged rather well:
stevepatterson.substack.com/…
"In my mind, the breakthrough comes from a revelation about the philosophy of language. This revelation is so incredible and counter-intuitive, that people find it easier to claim the machine has a soul than to update their philosophy of language.
In fact, I expect the only people who are nonplussed by the power of LLMs are those with a soft spot for occultism of some sort—those who think words are magical...
It turns out, a whole bunch of stuff is encoded in our language. So much, in fact, that by tracking patterns in our language, a mindless machine can effectively reason through concepts. That is, by mimicking our language, the machine can mimic our reason."