hamburga

> As he films them from multiple angles, the predominantly English-speaking employees ask the presenter questions such as, "What are you doing?" and "Are you filming me right now?"

Meta employees themselves couldn't tell if he was recording!

Chicago resident here.

Not disputing the methodology, but our state is the nation's leader in tornadoes (by more than a factor of 2) due to Niña/Niño this year, and Chicago spent a few days this summer with the world's worst air quality due to Canadian wildfire smoke.

Obviously, "resilient" ⍯ "unaffected".

As far as I can tell, there were no credible criminal charges against this group, except that they are effective organizers on the opposite side of political spectrum. In the White House’s deranged logic, this qualifies them to be officially designated as domestic terrorists [1].

Can we all agree that this is an assault on the First Amendment?

[1] https://www.whitehouse.gov/presidential-actions/2025/09/desi...

>> This raises an important question: if AI models can be misused for cyberattacks at this scale, why continue to develop and release them? The answer is

> Money.

For those who didn’t read, the actual response in the text was was:

“The answer is that the very abilities that allow Claude to be used in these attacks also make it crucial in cyber defense.”

Hideous AI-slop-weasel-worded passive-voice way of saying that reason to develop Claude is to protect us from Claude.

Socrates famously complained about literacy making us stupider in Phaedrus.

Which I believe still does have a large grain of truth.

These things can make us simultaneously dumber and smarter, depending on usage.

Yup, though he was also intentional about not copying word-for-word, but rather trying to predict the next token (or phrase).

https://muldoon.cloud/2025/05/17/frankin-llm.html

But assuming no new models are trained, this competitive effect drives down the profit margin on the current SOTA models to zero.

> This reminds me of a paradox: The AI industry is concerned with the alignment problem (how to make a super smart AI adhere to human values and goals) while failing to align between and within organizations and with the broader world. The bar they’ve set for themselves is simply too high for the performance they’re putting out.

My argument is that it’s our job as consumers to align the AIs to our values (which are not all the same) via selection pressure: https://muldoon.cloud/2025/05/22/alignment.html

I’m still waiting for somebody to explain to me how a model with a million+ parameters can ever be interpretable in a useful way. You can’t actually understand the model state, so you’re just making very coarse statistical associations between some parameters and some kinds of responses. Or relying on another AI (itself not interpretable) to do your interpretation for you. What am I missing?

There are a lot of people here who reflexively flag anything remotely close to US politics.

How long until Palantir gets a contract to do the same work for Russia?

AI is definitely a significant force multiplier in many areas. Still, an individual in total isolation has limited agency.

> The argument that persuaded many of us is that people have a lot of desires, i.e., the algorithmic complexity of human desires is at least dozens or hundreds of bits of information

I would really try to disentangle this.

1. I don't know what my desires are. 2. "Desire" itself is a vague word that can't be measured or quantified; where does my desire for "feeling at peace" get encoded in any hypothetical artificial mind? 3. People have different and opposing desires.

Therefore, Coherent Extrapolated Volition is not coherent to me.

This is kind of where I go when I say that any centralized, top-down "grand plan" for AI safety is a folly. On the other hand, we all contribute to Selection.

All it takes for somebody to nuke Atlanta is an atom bomb and an airplane and somebody willing to fly the plane.

I’m being facetious but there ARE ways to decide/act as a society and as subgroups within society that we want to disallow and punish and select out qualities of AIs that we think are unethical.

A single person acting in isolation (no friends, no colleagues, no customers) has very little agency. While theoretically a single person could release smallpox back into civilization, we have collectively selected it out very effectively.

Thank you for this. It gets exactly to the heart of the issue and what I sense is being missed in the AI alignment conversation. “What does ‘aligned’ mean” is and ethical/political question; and when people skip over that, it’s often to (1) smuggle in their own ethics and present them as universal, or (2) run away from the messy political questions and towards the safe but much narrower domain of technical research.

> AI Alignment is different because it is trying to align something which is completely human made.

Not sure what you’re getting at here; pharmaceuticals are also human made. The point in the blog post was that we should also want drugs (for example) to be aligned to our values.

> What I think is absolutely important to understand is that throughout human history "alignment" has never happened.

Agree with that. This is a journey, not a destination. It’s a practice, not a mathematical problem to be solved. With no end in sight. In the same way that “perfect ethics” will never be achieved.

Agreed - and this was definitely my intent with the blog post. If you only do Selection passively, you’re abdicating your ethical responsibilities to contribute to AI Alignment.

Why can’t we infuse Selection with Goodness? We’re the ones doing the selecting. We’ve selected out things like chattel slavery, for example.

(Disclaimer: fell asleep after 10 minutes of reading the SSC post last night. I know it’s part of the HN Canon and perhaps I’m missing something)

Indeed. Tolstoy does a deep exploration of this in War and Peace, using the example of Napoleon.

Of course, this gets to the heart of the free will debate (to be settled in a future post ;)). Both are true at the same time - organized people and dictators and other factors simultaneously wrestle for influence in complex ways in which causation is impossible to measure.

My own two cents, though, is that the Categorical Imperative is a tremendously important and underappreciated tool for raising the self-consciousness of groups.

A practical implementation of it is linked at the bottom of the blog post.

Also see Commandment 7 - https://muldoon.cloud/2023/10/29/ai-commandments.html

This is how society collapses and OpenAI wins.

Not totally following your last point, though I do totally agree that there is this historical drift from “AI alignment” referring to existential risk, to today, where any AI personality you don’t like is “unaligned.”

Still, “AI existential risk” is practically a different beast from “AI alignment,” and I’m trying to argue that the latter is not just for experts, but that it’s mostly a sociopolitical question of selection.

> this is a call for more people to enter the space

Part of my argument in the post is that we are in this space, even those of us who aren’t ML researchers, just by virtue of being part of the selection process that evaluates different AIs and decides when and where to apply them.

A bit more on that: https://muldoon.cloud/2023/10/29/ai-commandments.html

Thatcher was briefly a research chemist; it’s extra weird that she couldn’t see beyond the atomic unit, to the societal equivalents of complex organic compounds.

Agreed. I see this more and more as the AI safety discourse spills more into the general lexicon and into PR efforts. For example, the “sycophantic” GPT 4o was also described as “misaligned” as code for “unlikable.” In the meme, I filed this under “personality programming.” Very different from the kinds of problems the original AI alignment writers were focused on.

I find myself using the term “human superorganism” a lot these days. Our bodies are composed of many cells and microbes working together; and if we zoom out, we have the dynamics in which human individuals operate together as superorganisms. IMO there’s a whole world of productive thinking to be done to attend to the health and strength of these entities.

Fun read, thanks for posting!

> If I had time to do it and if I knew more about how AIs work, I’d be putting my energies into building AIs whose sole purpose was to predate upon existing AI models by using every conceivable strategy to feed bogus data into them, interrupt their power supplies, discourage investors, and otherwise interfere with their operations. Not out of malicious intent per se but just from a general belief that everything should have to compete, and that competition within a diverse ecosystem produces a healthier result in the long run than raising a potential superpredator in a hermetically sealed petri dish where its every need is catered to.

This sort of feels like cultivating antibiotic-resistant bacteria by trying to kill off every other kind of bacteria with antibiotics. I don't see this as necessarily a good thing to do.

I think we should be more interested in a kind of mutualist competition: how do we continuously marginalize the most parasitic species of AI?

Seems like a fellow traveler with the Wolfram Model of physics. But they’re also making falsifiable predictions. According to Gemini 2.5 Pro:

“Based on the content of the provided PDF, "RBT_v1.0_release.pdf," the research presented is not connected to the Wolfram Model. Regarding experimental results and falsifiable predictions, the document explicitly states that the theory of "Recursive Becoming" is falsifiable. It further outlines six concrete, parameter-free predictions that are intended to be testable before the year 2030: * Test 1: Ring-aperture X-ray line (XRISM) * Test 2: Axial-lepton phase shift (MAGIS-100) * Test 3: EDM sign (CASPEr) * Test 4: Missing-energy bump (Super-Charm) * Test 5: Curvature waveguide loss * Test 6: Immersive VR power budget”

We just need to have some art and theology in place by 2030 to make human sense of this :)

Blue Origin (https://www.blueorigin.com) | Software Engineer / Software Manager / Technical Product Manager | Seattle/Huntsville/Florida (No Remote) | Full Time

Blue Origin develops reusable rockets and engines (including the now-orbital New Glenn: https://arstechnica.com/features/2025/01/after-the-success-o...). We're also working on a human-rated lunar lander (https://en.wikipedia.org/wiki/Blue_Moon_(spacecraft)), a space station (https://en.wikipedia.org/wiki/Orbital_Reef), and some other neat projects; with goal of improving life on Earth.

My team is in supply chain and manufacturing technology. We're ramping up like crazy and have a super ambitious roadmap.

- Software Engineer I | $99-139k (WA, other locations vary): https://blueorigin.wd5.myworkdayjobs.com/BlueOrigin/job/Seat...

- Software Engineer II: https://blueorigin.wd5.myworkdayjobs.com/en-US/BlueOrigin/de...

- Software Engineer III | $148-207k: https://blueorigin.wd5.myworkdayjobs.com/BlueOrigin/job/Seat...

- Sr. Software Engineer | $177-248k: https://blueorigin.wd5.myworkdayjobs.com/BlueOrigin/job/Spac...

- Principal Software Engineer | $207-290k: https://blueorigin.wd5.myworkdayjobs.com/BlueOrigin/job/Hunt...

- Sr. Software Engineering Manager | $191-267k: https://blueorigin.wd5.myworkdayjobs.com/BlueOrigin/job/Spac...

- Sr. Technical Product Manager | $164-230k (WA, other locations vary): https://blueorigin.wd5.myworkdayjobs.com/en-US/BlueOrigin/jo...

If that's not your bag, we're hiring in just about every other engineering discipline, including lots of embedded software. 411 open positions as of this writing: https://www.blueorigin.com/careers/search

(Export control laws require us to hire exclusively US citizens, Green Card holders, or those in "protected person" status)

Feel free to email me with questions or thoughts! mmuldoon [at] blueorigin.com