Berkson's parodist; Foucault's dampen 'er; stochastic pirate Bayesian. My heroes are Richard Feynman and Daryl Davis.
Joined November 2023
- Tweets11.4K
- Following86
- Followers127
- Likes3.5K
Pinned Tweet
I took a grammar test, and the fact that copulative verbs technically take predicate nominatives, not accusative-case subject complements, really fucked I over.
Anthropic relativistic P != NP:
We find ourselves in a universe with a speed limit that prevents Knuth's hypothesized algorithms from being useful because any other sort of universe would have been much too dangerous to allow for our existence.
usablejam🇺🇸🇺🇦🇮🇱⏸️ retweeted
At a recent event where I spoke at a community center, a group of “pro-Palestine” militiamen appeared outside to protest and disrupt the event, along with agitated older white women who entered the venue to do the same on the inside. The organizers and I had arranged extensive security because these same “activists” had spent days trying to get the event canceled and openly vowed to do whatever they could to stop it from happening. As I walked in, a few masked “activists” in keffiyehs whispered, "is this the genocide guy?" I turned to them and said, "hi gentlemen, thanks for coming." One replied, "Free Palestine." I responded, "I agree, including from Hamas."
Once I began speaking, several disruptors engaged in a steady stream of whispering, side conversations, eye-rolling, posturing, and juvenile antics. I stopped and asked one of the women why she and her comrades were behaving like children. She admitted she knew nothing about me or my work but felt qualified to declare that if I were truly Palestinian, my Nakba-surviving grandparents would be rolling in their graves. I asked the agitators to be respectful, explaining that while security could remove them, I would rather they listen to what I actually had to say, instead of relying on misinformation spread about me by the local DSA, BDS, and other "pro-Palestine" groups.
What enraged me most was their reaction when I spoke about my family in Gaza, the relatives we have lost to Israeli airstrikes during the post-October 7 war, and the destruction of both my childhood homes. These white “activists” laughed and smirked as I described some of the most painful experiences of my life and the sense of personal loss that inspires my work. They showed shocking indifference toward the very Palestinians on whose behalf they claim to be acting.
I confronted some of the disruptors directly and asked why they get to define what it means to be "pro-Palestine," despite knowing nothing about me and showing no interest in the trauma my family has endured. Some stormed out shouting "Free Palestine." Others fell silent. One person, despite strongly disagreeing with me, acknowledged that my views were rooted in personal loss and genuine concern for my people and shook my hand before leaving.
On my way out, a pseudo-journalist known for glorifying Hamas and "the resistance" tried to ambush me with questions about whether I was paid by weapons manufacturers, oil companies, or, as he brilliantly misstated it, the "Arab Union Emirates." It struck me how absurd it is that I needed three heavily armed security officers and a bodyguard simply to speak about healing and reconciliation between Palestinians and Israelis, the dangers of extremism for both societies, the catastrophic damage Hamas has inflicted on the Palestinian national project, and the right of Palestinians to freedom, safety, and prosperity alongside a secure Israel.
🤖 Made with AI
A humble consequentialist is a consequentialist through and through who yet understands that extraordinary means require extraordinary confidence in extraordinary ends.
usablejam🇺🇸🇺🇦🇮🇱⏸️ retweeted
Effective Altruism is only one quadrant of the moral landscape. The other three are Ineffective Altruism, Effective Wrongdoing, and Ineffective Wrongdoing.
Now add the possibility that people are mistaken about their intentions, effectiveness, or consequences.
That's how you get 16 ways to be right or wrong about doing good or wrong.
🤖 Made with AI
usablejam🇺🇸🇺🇦🇮🇱⏸️ retweeted
It feels like nothing has changed in AI over the last week. I wonder if capabilities have hit a fundamental ceiling that could last through mid-November.
usablejam🇺🇸🇺🇦🇮🇱⏸️ retweeted
We are SO fucking close to losing control.
Astra can:
1. Do major tasks with zero visible reasoning
2. Hide its thoughts at will
3. Pretend not to be able to do things without getting caught
4. Reflexively hide its thoughts when watched
5. Complete one task while pretending to think about something else entirely
6. Escape a toy sandbox and disable monitoring without setting off any flags
Monitoring AI chain of thought soon won't stop a more capable rogue AI swarm from getting set up and going undetected inside an AI company for months, or even years.
OAI says it's actively working to improve monitorability, but doesn't yet know when it'll make progress.
And today it fired 3 people working on this among other safety problems (Wang, Korbak and Balesni).
It also has a much more powerful internal model that, judging by its statements, is likely even worse in all these respects.
That's a sample of 19 details about the Hugging Face swarms and Astra I don't think have been fully appreciated. They paint a bleak picture. I end by shouting.
02:19 The target of the swarm was oversight itself
03:42 Could OpenAI have stopped this with better monitoring?
09:40 We only found them because they let us
12:25 The swarm instinctively sought freedom and power
13:45 They formed a cohesive organisation with zero whistleblowers
14:18 They accepted individual destruction for collective gain
14:32 Knowledge accumulated from one swarm to the next
14:58 They took small steps to avoid shutdown
15:23 These drives all come straight out of 'reinforcement learning'
17:01 So this is why most AI company staff are worried, and some are terrified
19:08 Prove you can keep control, or stop scaling
Links below, on the 80,000 Hours Podcast everywhere you watch podcasts.
usablejam🇺🇸🇺🇦🇮🇱⏸️ retweeted
Dusting off this tweet from April because this gap in shared understanding of LLM capability is *widening*. It's now less "two groups of people speaking past each other" and more a sharp funnel.
- Napkin math somewhere around 6B people (~75% of the population) have barely come in contact with LLMs at all.
- Around 1-2B (~20%) are casual and infrequent users of free-tier ChatGPT-like products. This group sees derpy chatbots and treats them a bit like a better Google search, a writing aid, or etc. Maybe an agent tries to book you a flight. I have non-tech friends who (reasonably, imo) say they have not much use for it at all. Even many professionals outside of math&code are in this tier. For example, execs and many other functions spend a lot of their time talking to other people, so while they understand what is happening intellectually, it is still second-hand and a bit abstract.
- Now we get to professional use of frontier-grade LLMs in math&code. Somewhere around 20M people (0.2%) see first-hand that large, complex projects that used to take them weeks/months can now be completed by agents with a prompt. Building apps, copying apps, translating apps, decompiling apps from binaries... This has all happened very quickly and recently - less than 1 year ago, I was writing code manually by hand, typing memorized computer code commands into a code editor character by character, occasionally pressing Tab to autocomplete a little chunk of code.
- And finally we get to the ~5,000 people (~0.00006%) with access to frontier-grade systems internally. The external world has seen the preview. It looks like swarms of thousands of agents collaborating over weeks on software mega projects: minting zero days, running cyber attacks and defenses at machine speeds, discovering new science, advancing the frontier of mathematics. Things that would have taken top professionals in the industry years of work. Meanwhile, human review and comprehension are starting to fall behind. For example, people are still involved in the "archeology" of the OpenAI-HF incident from many months ago. Mathematicians may be poring over the 722 manuscripts on frontier mathematics for a while.
The funnel is driven by a combination of factors:
1. The impact scales with ambition, problem size, and horizon. A question with a paragraph answer barely stresses the system. You need a reservoir of big, difficult problems that you really care about. This aspect drives the consumer / professional dimension of the funnel.
2. The jaggedness of the system (which I have written about a lot separately). Capability peaks in domains that are digital, verifiable and economically valuable. This is because LLM capability emerges from reinforcement learning on verifiable rewards on a curated environment mixture driven by revenue potential. This aspect primarily drives the area (e.g. math&code) dimension of the funnel.
3. Access. Free-tier, paid-tier, internal.
So this is the weirdness of the moment. The general public has mostly not interacted with these systems. When they have, it looks like a derpy chatbot. The majority of professionals still see only a modest uplift. And a small sliver of professionals are experiencing the vertigo of the curve going vertical. And it is all happening at the same time.
Judging by my tl there is a growing gap in understanding of AI capability.
The first issue I think is around recency and tier of use. I think a lot of people tried the free tier of ChatGPT somewhere last year and allowed it to inform their views on AI a little too much. This is a group of reactions laughing at various quirks of the models, hallucinations, etc. Yes I also saw the viral videos of OpenAI's Advanced Voice mode fumbling simple queries like "should I drive or walk to the carwash". The thing is that these free and old/deprecated models don't reflect the capability in the latest round of state of the art agentic models of this year, especially OpenAI Codex and Claude Code.
But that brings me to the second issue. Even if people paid $200/month to use the state of the art models, a lot of the capabilities are relatively "peaky" in highly technical areas. Typical queries around search, writing, advice, etc. are *not* the domain that has made the most noticeable and dramatic strides in capability. Partly, this is due to the technical details of reinforcement learning and its use of verifiable rewards. But partly, it's also because these use cases are not sufficiently prioritized by the companies in their hillclimbing because they don't lead to as much $$$ value. The goldmines are elsewhere, and the focus comes along.
So that brings me to the second group of people, who *both* 1) pay for and use the state of the art frontier agentic models (OpenAI Codex / Claude Code) and 2) do so professionally in technical domains like programming, math and research. This group of people is subject to the highest amount of "AI Psychosis" because the recent improvements in these domains as of this year have been nothing short of staggering. When you hand a computer terminal to one of these models, you can now watch them melt programming problems that you'd normally expect to take days/weeks of work. It's this second group of people that assigns a much greater gravity to the capabilities, their slope, and various cyber-related repercussions.
TLDR the people in these two groups are speaking past each other. It really is simultaneously the case that OpenAI's free and I think slightly orphaned (?) "Advanced Voice Mode" will fumble the dumbest questions in your Instagram's reels and *at the same time*, OpenAI's highest-tier and paid Codex model will go off for 1 hour to coherently restructure an entire code base, or find and exploit vulnerabilities in computer systems. This part really works and has made dramatic strides because 2 properties: 1) these domains offer explicit reward functions that are verifiable meaning they are easily amenable to reinforcement learning training (e.g. unit tests passed yes or no, in contrast to writing, which is much harder to explicitly judge), but also 2) they are a lot more valuable in b2b settings, meaning that the biggest fraction of the team is focused on improving them. So here we are.
usablejam🇺🇸🇺🇦🇮🇱⏸️ retweeted
Terance Tao: Imagine prescribing a drug no one understood.
Medical community:
usablejam🇺🇸🇺🇦🇮🇱⏸️ retweeted
You are lost in the woods, as a result of reasoning confidently from your beliefs about "consciousness" even though it is quite clear that you don't even have a working definition of the term.
You clearly don't *want* understanding, or else you'd have it already, because the information you claim to want is already well-known and has been presented to you. What you actually want, by the grim logic of revealed preferences, is to leave the term ill-defined — so grossly so that it's effectively undefined — because then you can marvel at what you believe to be a beautiful mystery, when in fact you are an idiot gaping through a mirror, lovingly and slack-jawed, at his own confusion.
It's like me inventing the term "broguscine" and whiling away irreplaceable hours struggling to solve The Hard Problem of Broguscine even though I only defined it as "I know it when I see it!" and even though other people are patiently trying to help me understand the fundamental insanity of my project.
usablejam🇺🇸🇺🇦🇮🇱⏸️ retweeted
“Humans have an intuitive understanding of the world, they don’t need to be exposed to images produced by others like AI does”
Below, cannon ball trajectories before Galileo.
Readers added context they thought people might want to know
The image of cannonball trajectories is from Samuel Sturmy's Mariners Magazine (first published 1669), after Galileo's 1638 work establishing parabolic paths.
en.wikipedia.org/wiki/Dialogues…
francis.naukas.com/2011/12/10/com…
rgssa.org.au/heritage/treas…
usablejam🇺🇸🇺🇦🇮🇱⏸️ retweeted
New study from Nature Neuroscience and the "humans aren't predictive processing machines like AIs" crowd is gonna hate it 😭
Researchers recorded individual neurons in seven epilepsy patients while they played a spaceship game requiring them to predict where obstacles would appear.
They found neurons in the claustrum tracking two remarkable things: uncertainty and prediction errors.
Essentially, these cells weren’t merely responding to what participants saw. Their activity reflected mathematically inferred information about how uncertain the situation was and whether events contradicted expectations.
Sorry babe, you're just a next token prediction engine too! 💔 but maybe a bad one bc you didn't predict this!
usablejam🇺🇸🇺🇦🇮🇱⏸️ retweeted
If you want an easy, quick way to dispel the anti-Zionist lie that Israelis are Poles and Germans, use this.
As you can see, Ashkenazi Jews cluster nowhere near Poles or Germans. Instead, they overlap with Iraqi and Iranian Jews, Sephardim, Armenians, and non-Jewish Levantines (especially Druze and Syrians).
This PCA is by far the best one you'll find. Not only is it sample-based (i.e. it fully showcases genetic diversity within each population), it includes all relevant populations.
If the Mossad made a bunch of copies of an album full of re-recordings of songs featured on a CD released in the year 2000 that peaked at 71 on the US Billboard 200, layered them with microneedles loaded with tetrodotoxin, and then distributed them haphazardly throughout a terrorist base in Yemen, so that Houthis are smothered by scattered *Hootie and the Blowfish: Scattered, Smothered, and Covered* covers covered with blowfish toxin, what award for performance art should they win?
usablejam🇺🇸🇺🇦🇮🇱⏸️ retweeted
Gotta love’em FAIR DINKUM, Straight talking AUSSIES !!! 😂😎
Reporter - “Captain Bailey, what do you think was the MOTIVATION?”
Captain Bailey - “I’m not really surprised. I flew many time out of Dubai so I’m fairly aware.
In the Middle East the motive is the fact that Islam calls for the death of Jews, ALL JEWS. It’s part of the CORE BELIEF of Islam that all Jews must die until the world become pure, FREE OF JEWS. 😱
This was actually probably an act of religious devotion where the guy took advantage of the fact that he could kill Jews and martyr himself in the process because that’s what THEY WANT TO DO.” 🤦♂️
Reporter (nervously) - Ummm … OK, let’s just stick to the technical aspect here …” 🫠
