No, it might look like a collision or even pre-image attack, but it really isn't. It hinges on the idea of quine, a program which reproduces its own source (obviously without just reading the contents from the file it's running from). The interesting part here is that the AI (ab)used the HEIC decoder to run computations.
Just for fun, here's a Python program that prints the md5 hash of its own source code
I find this research odd because I don’t know what field of study to call this. It’s not computer science. It’s almost like behavioral psychology, but for AI agents instead of humans or animals. That said, I feel like these studies lack a lot of rigor. Or maybe it’s a lack of clear direction. Like, what are we trying to determine here? Because these things often involve advanced AI agents and complex harnesses, but then it sounds like they are given a very minimal prompt. What are we really studying then? Would the outcome be completely different with a different prompt? Can we generalize these results to other models? Different generations of models, different levels of thinking? What if we tried fine-tuning the model to this scenario? Is this fundamentally how any LLM would act, or is this some artifact of our current, generalized fine-tuning?
Reads to me like Proto-psycho-history. Attempting to model behavior mathematically, with the assumption that all human behavior can be modeled this way, and that these models can be useful for predicting human behavior.
Saw an article a quarter century ago in The Atlantic about simple population models as a way to explain more complex behavior. One was: randomly distribute blue and red cells across a 2D grid. Every frame, pick a random cell: if it has at least one color like it nearby, leave it. If not, move it to a random location. Over many generations, segregated color formed. The article used this to describe something about segregated communities forming. I offer no opinion about the credibility of the article, but it was fascinating to me as a baby nerd, and seems at least a little related to this article’s line of thinking.
Once you understand how complex the human brain is, with 86 billion neurons and 100 trillion synapses, formed over billions of years of evolution, comparing that to the machines we have built makes them seem silly and extremely simple in comparison.
Psycho-history is fiction. You're not going to be able to model a single human much less millions of humans and their environments.
> Psycho-history is fiction. You're not going to be able to model a single human much less millions of humans and their environments.
Of course it is fiction, but to be precise, within that fiction it was also stated that it's not possible to model a single human; the whole idea was that large societies of humans behave statistically in a predictable way, exactly because then non-predictability of any individual human becomes statistically irrelevant.
I often wonder how much of the human brain is actually dealing with "thinking", by that I mean navigating, understanding complex systems, and what is there to keep you alive, walking and breathing.
I think this is mostly just software benchmarking. The most science that can probably be extracted from this case study is probably a couple data points on where we are and where we can improve with AI agents. It seems like a decent task to optimize for. It might also work as a proof of concept for an economics study or theoretical political science.
I think large LLMs can be seen as a representation of humanity's collective thought processes... almost as if humanity were a single entity.
They have read pretty much everything we have ever written, good and bad, so they're kind of a mirror of us, as a species. Not a perfect one, but still.
Aliens could probably use a llm to get a crash course on what humans are actually like, and how to interact with them.
I would agree with this if it weren’t for RL. At the RL stage, developers implement their vision of what good model behavior should be (as Dario said, they’re implementing their own meta personality in Claude during RL).
Sociology is the right concept. It sits somewhere between system analysis and emergent human "sort of" sciences, like economics and sociology, where we can detect structures but humans are too complex to surface any platonic concepts with much confidence.
AI interaction may give us a cleaner way to test and analyze emergent properties that sit beneath so of the softer sciences, giving us a harder science that we had no way to investigate before.
It’s the same way psychology works on humans. Black box observational stuff.
You can’t generalize anything here because the psychology of the model can be affected by system prompt, reinforcement training, and training data. What you observe here can be wildly different if you change any of the underlying things I mentioned.
> I find this research odd because I don’t know what field of study to call this.
Probably multiple with very complex system dynamics, feedback loops: Cliodynamics, Cultural evolution, Cultural anthropology, Archaeology, Sociology Political science, Institutional economics, Urban planning, Complexity science, Social ecology
Computers science is just applied math, has nothing to do with the machine.
The term computer was coined to describe humans engaged in applied math jobs before the machines existed.
The models are matrices of values engaged in reducing options to the most desired outcome given inputs. That's software dev.
"Fundamentally" or "fine tune" is meaningless; no model spontaneously emerged through a membrane from another reality. They’re all fine tuned from nothing into a model or from a general model to a specialized one.
You're circumlocuting semantics to say something for social media, but the physical provenance is clear. They're not of mythical origin and it's still math relative to set of memory values.
One: the ontological status of the mathematics is not so certain
Two: most effects of LLMs result from applying what you describe to natural language. Without it the formalism would be meaningless, it is the insane fact that language can store knowledge and is itself a somewhat decentralised thoughtprocess (independent on the individuals that produce the language, since those individuals die, whilst language is preserved). Point 2 is of course the outcome of certain presuppositions on what laanguage is - which is the point: LLMs being in the question for a theory of language. Its research seems not only informationtheoretical but also linguistic in nature. A lot of the proposed terms of (one of the) parent comment(s) are the fields that somewhat are engaged with that intersection in a way. With the noteable omission of cognitive science and cybernetics/systemstheory (the latter of which might be in the list; not a 100)
There is zero obligation for anyone else to sit around and wait for ontological analysis; what fundamental forces of physics demand this?
You're not engaged in scientific inquiry.
You're circumlocuting scripture. Memorization and parroting. You come off like Luke Skywalker screaming about Jedi Sacred Texts! being forgotten.
Will lean on my applied physics degree studying in labs entropy in various materials. Hands on material experience adding, subtracting power sources and levels to systems of various physical materials measuring for change; everything is implicitly mathematical ratios.
The language is just what we use to communicate in meat space. It doesn't actually define anything but our social relationships to ground truth. I don't need to hear from someone in academia especially obsessed with the rhetoric to make sense of my lived experience.
I suggest thinking of it as psychology on a mathematical model of collective human works, and I’m currently curious to see the differences of models trained only on specific cultures
The label might come later, but cooperative or adverserial behavioral patterns matter deeply when deploying agents to act as weapons of defense or business builders.
In the magic system of DND that would be part of the school of Conjuration, whereas elementalmagic would be part of the school of evocation.
Have some conjuration engineering and conjuration injection (which woukd probably fall into the school of enchantment in the next edition SnM (Systems and Meaning)
I actually really disagree with this. If we want our representatives to reflect the will of the people, they need to have an accurate understanding of their electorate. When issues come up we need a way for the people to be heard, and polling is an important part of that.
This conditional is true, but there is a related one: If we want our representatives to outmaneuver the electorate on behalf of donors, they must have an accurate understanding of what their electorate is paying attention to, and what they are not.
Surfacing understanding of the electorates beliefs and desires is good but double edged. I think we can balance it out with yet other strategies and don't agree with GP's cynicism here but it's worth taking on directly.
Isn't there a pretty clear line between the following?
> I'll give you a gift card if you tell me how you're going to vote tomorrow
and:
> I want to represent you well, lets talk about what's important to you
I wouldn't propose we interfere with the latter. Or is the former truly useful for improving the quality of representation? It seems to me like it's really only good for improving ad targeting. Any representative worth their salt can find a way to get feedback from their constituency without requiring the people to participate in a cold call from a research company whose objectives are unknown.
I’m not necessarily disputing your point, but it’s very odd to downplay one source for potential bias, and then follow up with a study from the CATO Institute. Like, the CATO institute is openly a libertarian think tank founded by one of the Koch brothers.
I also think it’s unfair to look at just food prices. Housing and healthcare have bloomed massively.
Of course you’re right. Food isn’t the whole picture, and no think tank is perfect.
We should evaluate research by the research itself, rather than by the author’s subjectivity credibility alone. The is the point I was trying to make by providing alternative research and adding a little bit of nuance to the reply, but perhaps I should’ve been more forthright in addressing my own bias as well.
> We should evaluate research by the research itself
Why is this the argument when someone challenges your source, but when it comes to someone else's source your argument is:
> This is research from a labor advocacy group, with the explicit aim of the organization to advocate for this sort of thing. This doesn't mean the research is is automatically invalid, but it is bias worth mentioning.
I try to do this as well. It certainly feels better and makes me more confident that the code will make sense when it implements it. But I’ve found even with a very firm design an LLM will undermine it in the details. It will be convinced there is a particular way to do something, I will tell it not to do that and give it a complete alternative, and it will say it implemented it as written. And if you just look at the data structures and function signatures things look fine. But you look at the actual implementation details and it will try in some awkward way to implement things totally differently.
She was acquitted of several counts of wire fraud to defraud patients, but that does not mean harm was not found to customers. I haven’t been able to find the specific statute, but essentially the jury found that the prosecution didn’t prove all of the elements of that charge beyond a reasonable doubt. We’ll likely never know the details, jury deliberations are secret. But there was absolutely harm to patients. Balwani was also charged with patient fraud and WAS convicted on those counts.
But the question was how is this deceitful? Like, when someone says a documentary is deceitful, I assume there’s some important information the documentary didn’t include, or provided outright false information. I get that the documentary went in a direction you didn’t like. I just saw it this afternoon, and didn’t feel the same way, but I’m still processing it and I do find your opinion interesting. But unless you are saying this is, like, false advertising, I don’t see how that qualifies as deceitful.
It’s been a while since I studied complexity theory, but is that what Blums speed up theorem says? A quick google search suggests it states that there exist computable functions where there is no optimal program that computes it.
Like, there exist problems that have a lower bound of n time where algorithms can get arbitrarily close to O(n) complexity but there is no program that actually reaches it (there is one that runs in time O(n^{1.1}), one that runs in O(n^{1.01}), etc).
But that doesn’t mean that there aren’t hard walls, it just means there are hard walls where every program’s runtime is > that wall, rather than >= that wall. But maybe I’m misunderstanding, or there are some corollaries I’m missing?
As an aside, as someone who studied more on the pure math side, Blums speed up theorem is rather intuitive. It looks like there might be a relatively elegant proof using topology. Blums speed up theorem is basically just a statement that the set of programs that implement a particular computable function doesn’t satisfy the greatest lower bound property. This is also the case for, for example, the set of rational numbers: there exist subsets of the rational numbers with a lower bound but not a greatest lower bound (like the set of all positive rational numbers q where q^2 >= 2). I’m hand waving, but I think in that same vein you can show that the set of programs under the topology of ordering by runtime is completely disconnected.
This is true for the general criminal conspiracy law, but be aware that an “overt act” doesn’t have to be an illegal act, just some action in furtherance of that crime. That can be purchasing a weapon, or scouting out a location. There may also be other laws in play depending on the specifics
Just having saulpw's physical address is already furtherance of that possible crime.
The commenters here are cute little HNers who think they have found a loophole in the law. They are not the first ones innover their head.
Spoiler: the law is written in words, and those words aren't strictly executed like in a computer program, they are interpreted by actual humans who can see what you are trying to do and will stamp it out.
Yup. HNers (and so many normies) believe that there are well defined rules they can follow and be safe. Spoiler, there is one law in the real world and that is might makes right, and you have no might, so therefore you will never be right. China for its many flaws at least is honest about this.
reply