It may come as a surprise, but this phenomenon of "uninterpretable" circuits designed by algorithms is 30 years old now.
Adrian Thompson's research in the 90s evolved FPGAs that did signal analysis with bizarre features:
- A tiny number of cells (far fewer than expected)
- No clock, despite performing signal analysis
- FPGA cells that were logically disconnected, but when removed caused the device to stop working
Even then their approach was taking advantage of the physics in the FPGA. One can only imagine how effective this could be when applied to circuit design with the compute budget of a frontier lab.
Those "evolved" FPGAs weren't much of a mystery. They just used undefined parameters (inductive coupling, power supply rail glitching, etc) to achieved the trained outputs. They didn't work when the ambient temperature changed, or when moved to another example of the same FPGA.
Was going to say much the same. I recall one story about a genetic algorithm to make an oscillator with the fewest possible components, and it successfully did so by surprising the humans with a single wire, i.e. an antenna picking up nearby stray RF.
That is my favorite part of GA. Gradient free optimization but it turns out making a good fitness function is hard and like 70% of the time it just exploits some assumptions or gap you have in your theories. Really reveals the problem in different ways that traditional ML.
As someone who does a lot of genetic programming (like, old-school, without AI/LLMs, etc), I can confirm that the fitness function is very difficult to get right, especially if you are trying to evolve programs that have "adversarial fitness" -- you'd need to maintain a hall-of-fame, and that just makes the runs take _much_ longer, because, chances are, your fitness function is the bottle-neck.
So, it is very hands-off, but also very expensive, and it is never clear if optimizing the fitness function is worth it, because the fitness function itself may be insufficiently or incorrectly specified.
However, I do think that people should try, even with just a whiteboard or a notebook, to design a fitness-function, for their problem, as if they were going to try to evolve it, because (1) it forces them to explicate their correctness constraints, and (2) they may discover that the program that they are trying to write _is equivalent_ to the fitness function.
I'll give you an example for point 2. Many years ago, I had to parse a gnarly language, and I chose to do it via Chomsky Grammars (that automatically build a tree based on the grammar-spec). Chomsky Grammars are cool, in that they are basically just a state-machine, but they are incredibly difficult to debug: when they work, they might work incorrectly (malformed tree), and when they fail, they give no reason for failure (because even with a trace, you are trying to figure out which backtrack should not have happened). So, out of desperation, I started to consider using genetic programming to just evolve a correct Chomsky Grammar. It became clear that there are only 2 possible fitness functions (1) a function that tests a hand-picked input against a hand-crafted tree-output (which is vulnerable to over-fitting), and (2) a function that is not (well, is much less) vulnerable to over-fitting, but is effectively a pre-existing, correct grammar that can produce those trees.
If you are in situation 2, then the genetic programming is not necessary, unless you are trying to create an optimized (or obfuscated) parser, and even then the optimization may be overfit to the test-inputs (even if they are generated test-inputs from the grammar itself). If you are in situation 1, then you are better off re-evaluating your approach (I abandoned the Chomsky Grammar notation, and invented one that is much easier to understand and debug, without losing any of the expressiveness -- it also happens to be slower, but fast-and-broken is worthless compared to not-so-fast-and-works-fine).
One place where genetic programming has been consistently awesome, is in parameter-search style problems (e.g. your genome is a long list of floats, representing weights and/or anti-weights, and you need to find out which weights give you more fitness (or less error)). I hear good things about variable-neighborhood-search, but have yet to try it.
That sounds apocryphal but there was a noted paper describing a frequency discriminator implemented using a genetic algorithm and it ended up tied to the exact piece of silicon used to evolve it, with logic cells not connected to anything still changing the output.
That second paper is absolutely amazing, I’ve always heard this story and never bothered to find the source.
The section with oscilloscope traces showing the progression of the “designs” over time was extremely interesting - I’d love to see what the 10x10 grid of functions looked like at each snapshot.
The other side is Cognitive Radio [1] which also evolve the OTA protocols for cooperative diversity from IEEE 802.22 onwards. Now I can see AI, via a local SLM/NPU plus agentic GNURadio loops for new radio use cases. This is going to be much more wide spread in the upcoming 3GPP 6G releases in 2030.
GA’s optimize only combinatorial problems though — where you have discrete set of choices (~genes) for each variable, and therefore do not have a gradient
One great application of AI design is patent poisoning. Use AI to churn out masses of variant designs, make them publicly visible on a web site, and if future patents come out use any collisions to invalidate them or at least restrict their scope (generalization of a patent is limited by prior art.)
I’m reminded of lawyer Damien Riehl’s (performative) reaction to the Sam Smith infringement decision, back in 2019/2020. He and programmer Noah Rubin algorithmically generated every possible melody (within a certain combinatorial space, in MIDI format as I recall), and purported to release them under CC-0 license [0]. He went on to attract some attention and explain his argument at a regional TEDx event [1].
I seem to recall legal commentators reacting with an eyeroll—apparently judges split much finer hairs than these for a living—but it was a cute stunt.
For sure, but I suspect the law might look similarly dimly on the argument that “machines systematically generated all possibilities in the problem space” === “somebody already had this idea.” I’d imagine maybe by reading specific human intention into “prior art” and “existing work” and those sorts of terms.
Which is not to say let’s not do it anyway and see!
There is a fairly effective counterpoint here that information is worthless if nobody can find it. Generating a handful of useful designs in a sea of pointless ones doesn’t count as ‘coming up with it’. Remember that a description of every idea already exists in the Library of Babel. I do not think that AI generation without curation really counts as ideation.
My point was that it’s hard to imagine citing something that could not be patented as prior art. It would be like citing a phone book as proof that a software program can’t be copyrighted (“the exact bytes appear in the 1973 Albany NY white pages, therefore it wasn’t original”)
"Humans couldn't even imagine" seems like overselling it, but I'm sure that machine learning algorithms can brute force their way to chip designs no one has tried before and that some of those might be useful to us. That seems like a pretty reasonable thing for a computer to do.
It's marketing bullshit. For one, it's like proving a negative; you can't prove to me that humans couldn't have imagined it. Second, humans have already imagined quite a lot of crazy stuff...
I wonder if our common expectation that true theories somehow had to be beautiful and elegant is going to survive the coming century. What if "real" nature phenomenon were actually best described by horrible mess of impossible equations, that only machines could actually manipulate and reason about ?
I think you're going too far with this. Most people understand scientific theories to be an approximation. F=ma is approximately true, in the sense that it's only accurate within the newtonian regime and each of those terms includes so many asterisks that you will only ever measure it approximately.
The latter is the jokes about the physicists "assuming a perfectly spherical cow."
In fact that's kinda the whole point of the "unreasonable effectiveness of mathematics" essay. It is unreasonable that mathematical approximations are so good at describing our world.
I often think this about medicine and the human body. We want to believe that our bodies are some miraculous well oiled machine. But it often seems that it’s a barely held together bag of mess.
Biology is incredibly robust!! I'd say barely held together bag of mess describes something like an internal combustion engine. A primate, on the other hand, is a self-replicating machine capable of self-repair and just about universal fuel sourcing. It has a robust defense network capable of identifying and eradicating a staggering number of foreign replicators. It has holographic design storage, with each cell containing the plan for the whole organisim. It has general cognition based on a world model, and does all this on almost no energy.
I think your reply and the parent can both be true, you're just using slightly the same words to describe different things.
The parent is talking more about elegant simplicity vs. sprawling, seemingly haphazard complexity, and you're talking more about durability to failure points and 'completeness'.
Likewise, in code, a lot of the most durable, battle tested software looks extremely inelegant and duct taped, as 90% of the code is dedicated to handling one-off patches and weird edge cases.
My suspicion is that we had a sense that generality and compactness was really neat, so we liked easily-remembered laws like F=ma. Applies everywhere, is clean.
When you attempt to hyper-optimize, even with humans in the loop, you end up a mess. You're lucky if you can find clean guiding principles anywhere. If you can hyper-optimize hyper quickly, you end up with an extra layer of mess.
But those proofs are showing that the fundamental axioms (which are generally simple and elegant) are still enough to build a complex result.
I think of elegance as not having to add epicycles, not that everything in the system has to be simple.
Also, without a working theory the, the space of possible solutions is near infinite. LLMs manage to pluck out the space of comprehensible English strings from n-dimensional hell. Even if this is done with a black box of billions of parameters, it’s still elegance in the sense that such a space even exists and was found
Math is a language to explain systems. Teaching someone that force varies linearly to mass is a helpful first pass. It isn’t exactly linear but is not exponential at all.
Gaining expertise is always the hard part and our new LLM overlords are making that much harder. So the simple “pure” functions as a teaching aid have never been more important.
End users have never cared about how the sausage is made though.
LLMs can explain complex things to humans with tons of specific context that you don’t find in textbooks or even a google search.
It’s probably never been easier to grasp a large codebase than it is today for example. You can probe and ask specific questions without going through a maze of imports and relationships and config files yourself.
Learning things will always be up to the person, it’s still a choice and dedication to a craft can still be taught.
The "common expectation" I think, misses the point. The idea isn't that fundamental theories are simple or elegant (quantum physics equations are pretty darn ugly), it's that, given the choice between a more complicated and a more simple theory, generally the simplest one is the most accurate choice.
Are you thinking of any specific examples? I don't disagree that complex things generally end up having complex explanations, but I'm admittedly drawing a blank trying to come up with things where the most complex explanation ended up being the correct one.
There are many marvels of evolution's ability to come up with robust complex distributed systems which work way better than anything we build. The one I've learned about most recently is the https://en.wikipedia.org/wiki/Immunological_synapse in which different kinds of white blood cells gather around a bit of evidence that one of them found and decide whether to shoot the messenger (clonal anergy), or raise a clone army to defeat the invader (T-cell activation).
Imagine that it's maybe the 1800's and you're asking why somebody who has already survived smallpox is not susceptible to becoming infected again. If you offered an explanation involving tiny detectives wandering around and collecting evidence which they present to each other and decide whether to multiply... one in which the tolerance comes from the detectives from the previous fight still hanging around in your lymph nodes ready to spring into action if they run across the right kind of evidence. Well that would probably be a more complicated explanation that anybody at the time would offer, and it would also be correct.
Incomplete prior knowledge doesn't mean it's simpler, just that it's inaccurate. Would the phenomenon you're describing really accurately be explained by something _simpler_?
You should look into Solomonoff induction. Nature and physics, absolutely, tautologically, have to follow the "shortest explanation is more likely principle".
Having theories that only give answers, but you can't reason about is not as useful. Having a theory where you don't know the limits of it's applicability, can be very dangerous.
At least in the physical realm there is not yet anything that combines relativity with QM so they can only be approximations. Even in math so far there seem to be similar challenges using programatic and "AI" driven solutions and proofs.
Still, I know that LLMs will be useful for Verilog/VHDL and particularly with verification, where they are already heavily used. Defined outputs and complete test coverage is already such a big part digital/asic design, I'd be surprised if it isn't used a lot more. Many software people would say that hardware is badly written copy-pasta, as it is. That said, higher velocity slop and hardware "technical debt" isn't something you can fix with an update. And no matter how fast you "ship", you won't get parts back in less than a few months. Poorly used, it will lead to expensive failures.
Solomonoff induction doesn’t concern itself with what is truth and reality. It just says which theory to prefer and how to determine so objectively when multiple are equally precise in making predictions of observations. It’s a formal description of Occam’s razor.
OPs argument is that reality is expressed by very complex equations and interactions; by definition this is outside of Solomonoff induction because it’s easy to imagine this accurate model by definition is the shortest algorithmic explanation, it’s just orders of magnitude more complex than our current approximations.
You should include the error correction code length in the description length. This means Newtonian mechanics was a much longer theory to describe Mercury's orbit than general relativity. It was only the shorter theory before they had the data showing a discrepancy. Which is the correct approach to describing your reality, because until you see a discrepancy, the extensional properties all follow the shorter rules.
I guess the argument from OP would look like: "Yes, now imagine we poke and extend our universe as far as we can. How much bigger do you think our final 'shortest description' would be? I imagine it may be orders of magnitude more complex."
Well, I can imagine a squared circle... doesn't mean the math checks out. I would reply that you do not have to imagine, you can go about looking at different mathematically possible universes in Tegmark IV and find the expected number of bits for the one you actually exist in. Which is ~0 bits more complex than the shortest description based on the data you currently have.
Also, note that Newtonian mechanics is not actually a very short theory for building a universe, because you have to instantiate every object in the universe. You actually get a lot more of the structure for free with general relativity (re: Wigner's classification of the particles). An observer in a presumed-Newtonian universe calling it a simple theory would be like saying, "I compressed Wikipedia to one byte, just by putting it all in the decompiler!"
>I wonder if our common expectation that true theories somehow had to be beautiful and elegant is going to survive the coming century.
That's the layman's idea of physics theories. They are beautiful and elegant only on the surface, that's why they're technically models and approximations of the real world. The standard model renormalization techniques are a mess of patches and ad-hoc heuristics, pretty far from the "this lagrangian literally contains all physics". Generally you just _ignore_ higher order terms and just call it a day. The famous E=mc^2 it's just the first term of a Taylor expansion. The beautiful form of physics it's what you would call "good enough" and often just a pedagogical tool.
It would represent a pretty sharp inversion from all the progress of mathematical physics until the present.
Up until the present it has been a nearly uniform march of revealed symmetries, collapsed privileged frames of reference, and other such (in the deepest sense) simplifications in our model of reality that has improved its fidelity to the measurable.
I hang qualifier about these developments being simplifying because the result isn't simple in the details: quantum chromodynamics is a daunting subject! But it's not just an enumeration of details and contradictions, the particle zoo that preceded the Eightfold Way looked like line noise, now in indexed notation the Lagrangian of the entire Standard Model fits on a page (or so I've been told I've never actually seen the page).
It's almost tautological that the frontier where it's still messy involves an unrevealed symmetry or a persistent privileged frame of reference, that's what frontier means, we don't see past it to the seam where it folds up.
Personally I suspect AI systems will be a great deal more inclined to discard the parochial axioms that have every point placed human ego above simplicity.
It doesn't resolve all of the open problems in physics if you amputate consciousness, free will, agency persistent identity, and an unambiguous arrow of time.
It would be really cool. We already know everything at the lowest levels is a probability cloud. There’s beauty and contentment in not really being able to nail anything down for eternity…
That's a result of the Copenhagen Interpretation. There are other interpretations of the math which don't rely on reality fundamentally being a probability cloud/wave/field.
How would nature be best described as a horrible mess of impossible equations? They would be best described as elegant and beautiful no?
I think your point is more that we might be able to initially describe complex phenomena as messy, horrible complex equations, that doesn’t mean we shouldn’t work to simplify them and make them more understandable to us.
I’m a bit frustrated. AI can do a looot of things; but I think as we continue to muddy the waters between LLMs and more traditional machine learning like Monte Carlo, Genetic Algoriths, Expert Systems and other Statistics magic tricks, we’re too aggressively conflating established and morally neutral activities in ML with the concerns that people have about LLMs and Stable Diffusion.
It is a problem because people will talk about what AI can do implying that an LLM can do that thing, making it seem like a pure LLM can do almost anything. On the other hand people will say AI will never be able to do X because an LLM can’t do that thing well natively. AI has become too vague of a term to be useful.
We're learning that people are way too lax with where they apply the term "intelligent". LLMs aren't remotely intelligent, but people are trying to ride the hype train and call them intelligence.
Maybe I'm just significantly and unrepresentatively unlucky, but Claude is significantly more intelligent than the average human around me on most any metric I can think of.
I wish I could wave a magic wand and just make the word "AI" go away. It has no actual meaning. It could mean anything from your opponent in Mario Kart to Stable Diffusion.
"AI" == "what (through tech) can replace a professional"
It may seem similarly vague, but it does in fact open interesting, productive, and necessary questions. A "computer" was a professional crunching numbers - "replaced", "easily" because of the deterministic procedural nature of said work, but what about the technical effort to arrive there, and what about the less "mechanical" jobs? When do "processes" become "intelligence"?
Some of us had studied AI originally to study the mind - "how do we formalize thought". It's the interdisciplinary, transversal nature of the area.
Also maybe compare that with that large and important intersection between CS and Economics - the "science of optimization" and its implementation in efficient IT systems. The effort in terms of that different discipline may not be evident, yet lots of engineering is "optimizing" and the generalization of those solutions we call Economics (see the book Algorithms to live by).
So: the term "Artificial Intelligence" may not be important as CS solutions to practical problems are built (you just focus on the better solution), but there is relevance to the "side disciplince" of AI, and from that perspective that is the cone, the scope anyway. "How would an intelligent solver approach the problem".
If what it does is "taking care of the carry", it represents a pretty minimal requirement for intelligence - it does replace a professional that could do it, but that professional does not have to apply too much proficiency and cleverness to do its job. It is improper AI.
> what type of software isn't AI
That which would not correspond to the job of an intelligent entity. Maybe blitting bitmaps around a screen?
As I tried to convey, it is more of a matter of perspective: the area of "implementing ways to solve problems as an intelligent entity would". It is a discipline that intersects others - engineering, logic, brain science, philosophy, epistemology, maybe again economics (as "the science of optimality and efficiency" - as an intelligent solver would do)... Consider it a special discipline that spans many other realms.
> If what it does is "taking care of the carry", it represents a pretty minimal requirement for intelligence - it does replace a professional that could do it, but that professional does not have to apply too much proficiency and cleverness to do its job. It is improper AI.
I think you're underselling how much mental work is required to solve complex arithmetic. Yes, it's simple for a computer, but (1) even basic computers are extremely complex in absolute terms, and (2) even the most complex computing tasks could be considered simple once you break them down far enough—for example, a large language model is "just" fancy matrix multiplication.
So I feel like there's a "sufficiently advanced technology is indistinguishable from magic" element here. Something becomes AI once it seems sufficiently advanced. But then time passes and it doesn't feel that advanced anymore.
I understand that human language doesn't always have a super precise definition, and I'm not trying to be pedantic. I think the term "artificial intelligence" is under-specified to the point of having virtually no meaning. To the extent that it is useful—obviously, a lot of people are using it conversation, so something is getting communicated—it's because it's possible to infer from context what someone is referring to (ie "the student used AI to write her essay" is clearly referring to an LLM, not Eliza).
We'd all be better off if we used words that describe what we're actually talking about.
Game AI uses behaviour trees, usually coded by hand. Decision trees are used for classification and are normally learned from data. The latter are a traditional AI technique from the early days of the modern machine learning era, in the 1990's.
I disagree. AI is doing exactly what it was predicted it would in science fiction.
The computer can now literally talk to you in natural language and then perfectly produce sophisticated actions in response to completely arbitrary and unstructured input. It trivially passes the Turing test. By any definition prior to the year 2023 we are living with Artificial General Intelligence and it’s here now.
Remember, the interrogator is allowed to be hostile, so they would obviously employ all known prompt injections and typical LLM 'gotchas' to figure out who the AI is.
So where are the androids? If it's AGI, why is it used as a tool, waiting to be prompted or executed by humans? Where is Skynet? Military applications still rely on human operators.
Yes, but unlike a lot of science fiction, robots, LLMs and other AI remain tools for human use. Augmented Intelligence would have been the more accurate word for real world AI.
I doubt LLMs will give us full embodied intelligence that science fiction androids have. Maybe some other approach. But I suspect for the forseable future LLMs, robots and other AI methods will remain tools, not independent agents like Star Trek Data or Skynet.
Yeah. We'll be arguing "is it really AGI" for many more years. Meanwhile, everyone interesting is going to have moved on from that question, choosing to spend time on "who cares if it's AGI, can it do $foo", for whatever value of $foo is interesting to them. Whether the machine is folding clothes or folding proteins, AGI isn't well defined other than "I'll know it when I see it", so whether or not it's AGI, the question is what job is the machine capable of and is it cheaper than a human? A humanoid robot that can work a warehouse is not putting anyone out of a job if it costs a billion dollars, and neither is a digital AI employee that costs the company a billion dollars either.
"AI" is a term cursed by cool sci-fi implications. It makes it a kick ass marketing term because most people are going to have some familiarity with sci-fi AI and "X media predicted Y technology" is a pretty widespread belief for a lot of values of X (star trek, Hitchhikers Guide to the Galaxy, Arthur C. Clarke) and Y (internet, cell phones, VR). If you want to tell someone we're making big strides in something, linking it into some popsci understanding of sci-fi being the great predictor of human achievement is low effort and high impact for quite a few people.
People aren't trying to communicate accurately if their first priority is getting you excited about the thing!
I have been practicing saying ML for traditional machine learning and LLMs for LLMs for just this reason. Trying not to say AI anymore. Too ambiguous. Sometimes I'm talking about game AI even, I'll try to use shorthand for whatever algorithm I think the AI is using (often I'll talk about its flowchart, though not always sure it's literally using that under the hood).
What is ChatGPT then? Sure it's an LLM, but I can give the app pictures and audio, and it can generate pictures for me. Do we distinguish between the bits of the architecture to accomplish those features separately from the LLM part of the product?
Recently I heard some people conflate procedural generation and generative AI and I had to explain why there isn't some legal or ethical issue with what breaks down to essentially scattering some points.
It's really getting annoying having to have these conversations.
Just as more successful machine learning fields distanced themselves from the term during the AI winter, I suppose we will (and perhaps are?) be seeing them adopt it again, now that we are in an "AI summer".
As always, it's a matter of funding. Both inside academia and outside of it. I remember when nanotechnology was all the rage. Everyone flocked to writing grant proposals about their "nano" technology that was thousands or millons of nanometers, aka micrometers or even millimeters. Stupid but if it works it's not stupid. The old joke is what do you call AI that works? Machine Learning.
The real question is how much compute do you need. With LLMs getting popular, so is compute. That's the real win for non-LLM technologies. The sheer availability of GPU capacity. Yes, it's expensive, but time in a GB300 supercomputer isn't even possible if they don't exist.
Alexnet succeeded for many reasons but a big reason is that computers got good enough to apply those algorithms and techniques in practice. Outside of LLMs, what new AI/ML systems await us in the future? The LLM bubble popping, if it ever does, is going to leave us with supercomputer capacity going unused and available for cheap, meaning experiments that were once infeasibly expensive become practical. I can't afford $10 million to run a weather simulation, but at $1,000 for the same amount of compute, a lot more experimentation becomes practical.
AI is not a real thing or a natural kind but a perspective. Whether something qualifies as "AI" or not cannot be decided by the objective features of the thing. Ergo, it can be defined at the author's pleasure.
> conflating established and morally neutral activities in ML
LLMs are no more or less morally neutral than other ML techniques.
things like the ones I had enumerated in my first comment. I'm old enough to remember FUD about all of those and many others, and perceptive enough to identify many ongoing public opinion campaigns as such.
the biggest question for me is how robust are these designs.
in the journal articles they did show measurements of real devices which agreed fine with predictions, but i didn't find them addressing it explicitly in the text. also, some systems they presented contained subblocks that were conventionally designed that could be carrying some of the weight.
or maybe i'm just sour that they're coming for my job? or maybe that's what they want us to think?
i think what wins in practice is simple ideas that can work in spite of all manufacturing and environment variations, and model limitations -- think stuff like feedback and symmetry. and what they show here is the opposite of that.
i've done blind optimization of circuit parameters some times only to end up realizing some pretty simple such ideas that i'd missed (like "you need symmetry here" or "you just need more bandwidth here") and made complete sense when you thought about them. so i wonder if we can't tweak a few pixels in their structures and reveal something simpler.
> How generalizable are these methods? Can they consistently deliver truly high performance? Can we get to a place where AI produces designs that maximize every conceivable trade-off, holistically optimizing every parameter to its most ideal physical state? .... AI can hallucinate a design that creates bad circuits that don’t work. This means verification methods need to remain under human oversight.
And they are essentially correct. We need better validation and verification methods, both software and hardware to keep in check the mistakes of automated random processes.
> but i didn't find them addressing it explicitly in the text
Yes, this is exactly what bothers me about this article and about a few similar articles published in the past, that they do not contain any evidence that their claims about the usefulness of AI in design are true.
In TFA it says that the role of AI is replacing the electromagnetic simulator in the optimization process, by guessing the behavior of the structure, which is many orders of magnitude faster than a simulation.
This sounds plausible, but in order to believe this I would want to see the differences between AI guesses and real measurements, in the case of structures with geometries that are very different from those used in the training of the AI.
Also I would want to see exactly with which simulators they have compared the speed of the AI model.
There are various simulation approaches for electromagnetic fields and electronic circuits, that can trade-off accuracy for speed, so I am not convinced that AI inference takes necessarily much less time than some faster low-accuracy methods of simulation, which would still be more accurate and more reliable than AI guesses.
Since you beat me to it, I'll add something that relates relates you were saying on "realizing some pretty simple... ideas".
I think a big plus of computer aided design like this is "innovization"[1]. Somewhat awkward term. But, a system like this leading one to deeper understanding of a particular process is the general idea. It's a fun feeling in practice.
> the biggest question for me is how robust are these designs.
Maybe it doesn't matter?
I mean, of course it matters. But most of this sort of design space is effectively NP-complete, where the creation starts with a blank schematic page and has an impossibly large search space, but where the checking of the design is much simpler.
> also, obligatory mention: "genetic antennas"
Exactly. How does this work? When confronted with the question, of course, everybody gets all excited about the constrained randomness of the GA, but if you think about it, what really makes it work is that there is a comparatively cheap test for fitness for purpose.
I was going to post this as well, its delightful to see that other people enjoy it since it was really mind-blowing when I read it.
It's interesting since I saw another comment near yours that raised the question of robustness of the lab-grown design, which I thought was kind of the most fascinating part of the damninteresting article was the revelation that the evolved programs were inseparable from the single physical FPGA used in the training. Since this RFIC training model employs a simulator, do you suppose that the quirks of the physical hardware on which the simulator runs are sufficiently isolated from the training such that a pair of designs would behave similarly when the simulator was run on distinct hardware? And I guess the even more obvious question is whether a design evolved on a simulator would have any hope of behaving as expected in physical hardware?
My hunch about the latter is no, although it still seems like an interesting study, and I often find myself thinking that really understanding what was going on with the FPGAs might be a prerequisite for really understanding how to master reinforcement learning.
Anyway I'm glad you posted this and if you have any other favorites related to this domain send them my way!
One takeaway from the article is that they had to get rid of the tried and tested fundamental building blocks of chip design to generate this advancement. I wonder if the same applies for mundane coding. Are the incredible innovations in AI coding actually hampered by rust and python? Should we let AI tools just code in the lowest level possible?
No. I'm really not fan of Python but you're not going to parse JSON at assembler level. You need to choose right level of indirection for each task. And LLMs are good at more than Python and Rust. I'm using LLMs to write programs in my own language that is compiled by a very fringe language and it happily does everything from scripting on ESP32 to audio plugins and 3D gfx on desktop.
It's not really that magical. As TFA points out, RFIC design, way beyond normal RF engineering, is close to black magic that relies a lot on the knowledge and experience of the designer, assisted by what would have been supercomputer-level-a-few-decades-ago modelling and design tools. What AI can do is a breadth-first exploration of all possible outcomes and then pick the best-performing one rather than the human-level "this seems like a good path to go down, let's explore it further".
> But freed from the constraints of human-designed templates and the need for humans to even understand the rationale of electromagnetic structures, ... ICs ... can take on truly wild-looking yet efficient designs.
I feel like technology is going to become alien at some point. We're all going to be using magical runes instead of chips.
> That’s not even to speak of all the movie plots that would have been ruined.
I clicked on all the links. Pretty much all of those movies could still work with wired technology. Even the one called cellular, in which a woman is trapped in an attic with a broken landline phone and manages to connect wires and dial a random number.
Yes I'm nitpicking. I guess I'm glad we have Wi-Fi and all, but don't try to sell me on it as a crucial plot device
In case anyone feels déjà vu, Popular Mechanics wrote about this professor's lab in Jan 2025, with almost the same title: "AI Designed Computer Chips That the Human Mind Can't Understand".
I feel a bit of unease when I read this title, not because of the threat of AI, but because the prevailing aphorism that "RF is black magic" is a slap in the face to the millions of physicists and RF engineers who DO understand every bit of this. It's a fun harmless anti-intellectual saw that I don't believe is harmless at all. We need more RF engineers and telling people it's all "black magic" and "wizardry" (and worst of all, saying "even RF engineers don't understand RF") makes it seem like it's not worth studying.
> telling people it's all "black magic" and "wizardry" (and worst of all, saying "even RF engineers don't understand RF") makes it seem like it's not worth studying.
I think the opposite is true. It being advertised as difficult to understand is one of the reasons I personally decided to study RF Engineering. The prospect of learning something so challenging pulled me in. The Smith Chart helped.
Yep, can't wait until everything is free and costs nothing to generate content. Free hosting and electricity will be super sweet too. Won't need admins or even the Internet. Everything I want will just be free for me because I don't think anything has value.
The methods outlined in this article aren't new. Scientists were using "genetic algorithms" to design antennas that weren't understood by anyone, but worked well, decades ago.
Chips? I've tried to task Opus, Gemini and Codex with a simple PCB. All of them placed holes correctly but can't understand that the traces should not cross physically.
>To arrive at the “sweet spot” where all these different parameters are balanced into optimal harmony, designers will typically lay out several different versions of the circuit, using intuitions and methods they have picked up in their years of training.
I kind of thought the real success is when the designer comes up with key things that are well beyond their training or any training that could have been done up until that time. Based on their years of experience living in an environment where training is table stakes but that's not the thing that's relied upon the most in the end.
With LLMs it seems like odds are that a concept which is statistically insignificant in the training set may surface in place of a truly novel solution, effectively displacing the real breakthroughs that actually go beyond trainable performance.
In a way that decision-makers can not tell the difference, and that could be the worst part.
Utterly haphazardly and inconsistently of course, same as yours. You thought that was some sort of argument? It slots right in and contradicts nothing.
I am confused, every day I read on HN that AI's can just interpolate the data they have seen in training, and that they are structurally incapable of coming up with something new, creative and not in the training distribution.
This is analogies to finding a new prime number by brute force using existing maths, rather than inventing new maths to get there.
The AI in this case didn't create a novel technology- it merely used the existing technology without basing the new design on a previous one. The whole "human couldn't come up with it" is because the possible design space is so large, there's no reason a human would start where the AI did.
The thing the AI did better than humans was brute forcing a solution faster. Still a very handy thing to have, but it isn't "creating" in the sense that it invented new materials or fabrication processes or anything novel.
> I read on HN that AI's can just interpolate the data they have seen in training
No. That can be said about LLMs, but not about all forms of AI. The technique used is not a LLM.
Sadly we've bastardized the term AI that, if it ever meant anything, it's meaningless now. The currently most voted thread in this post discuses the topic.
Have you read the article? The creative element came from the researchers:
> In our new approach, the architecture begins essentially from nothing and is progressively assembled through successive iterations. The system explores the design space by generating myriad candidate circuit combinations and mapping the resulting performance trade-offs as it navigates this landscape. Because the process is not biased by prior human design choices, it can produce completely novel circuit topologies that look markedly different from those created by human designers.
In my experience, if you tell them to research the web to see if their idea has been pursued before, you can get them to keep proposing new things until something is sufficiently new, even if it's a new interpolation between existing concepts, that it's effectively an original idea.
This is wrong - the training data is necessary but insufficient. There are a lot of other parts of the architectures used that add a lot of value - otherwise Markov chains would be all you need. There are layers upon layers with non linear activation functions, learned residuals, etc. They still absolutely must interpolate but the space they interpolate through is much more complex than the training data, and they can definitely create things not in their training data. What they can not do is wander outside their non linear parameter space’s convex hull. But this is a really permissive constraint on what they can do “creatively.” People generally under estimate the advantage the architectures confer on that constraint. This is why there was a step function change in expressive power as the architectures (attention, self attention, transformers, diffusions, others) evolved given the same training data. Generally though I challenge you to define “creative” in a way that is precise enough to measure and isn’t self referential or refer to concepts ill defined.
The key tho is can they solve problems not easily solved before with prior techniques. Further can they identify problems not readily presented. Then identify novel solutions. Etc. The answer is emphatically yes they can. These features don’t have to literally exist in their training data, but the supporting highly convoluted network of associations of all their training data does have to in some complex space allow for it to produce these answers. It’s not the same as they’re stochastic parrots at all.
Are they creative? No, because they don’t have awareness. My personal imprecise definition of creative requires both self and awareness as well as free will. There is no driving awareness in all AI architectures, it all derives from extrinsic impetus. Creativity is derived, IMO, from a layer of our minds that is not readily assessed or measured and is only indirectly expressed through language, art, and music. Hence it is not directly trainable and therefore a learning model can’t learn it by reinforcement. It can learn the proxies, but the proxies are not, as we all deeply know, the same as our experienced awareness. We are not our words, our art, our music. We try hard to bridge it, but it’s impossible and you and I know this to be true from experience. In fact we can not even examine our own awareness because it’s not directly observable or possible for us to directly reason about. This is core to a lot of philosophy, especially mid and far eastern philosophy of the mind, the self, the five aggregates of Buddhism, etc. Psychology points at it, and modern psychology avoids it because it’s practically difficult for outcome oriented treatments.
>Generally though I challenge you to define “creative” in a way that is precise enough to measure and isn’t self referential or refer to concepts ill defined.
While I have no hope for a rigorous definition (I don't think it's possible), there are two very distinct kinds of creativity:
1. Result is sufficiently novel for the system itself, i.e. it never seen it previously. This kind is too trivial to even talk about.
2. Result is novel for the side observer. This kind of creativity is meaningless because it depends on at least one unknown (side observer).
It may come as a surprise, but this phenomenon of "uninterpretable" circuits designed by algorithms is 30 years old now.
Adrian Thompson's research in the 90s evolved FPGAs that did signal analysis with bizarre features:
- A tiny number of cells (far fewer than expected)
- No clock, despite performing signal analysis
- FPGA cells that were logically disconnected, but when removed caused the device to stop working
Even then their approach was taking advantage of the physics in the FPGA. One can only imagine how effective this could be when applied to circuit design with the compute budget of a frontier lab.
https://cacm.acm.org/research/analysis-of-unconventional-evo...
Those "evolved" FPGAs weren't much of a mystery. They just used undefined parameters (inductive coupling, power supply rail glitching, etc) to achieved the trained outputs. They didn't work when the ambient temperature changed, or when moved to another example of the same FPGA.
[flagged]
Reminds me of good ol genetic algorithm search. Guess and check can be quite powerful, especially if you can toss in agent in the loop guidance.
https://en.wikipedia.org/wiki/Evolved_antenna
Was going to say much the same. I recall one story about a genetic algorithm to make an oscillator with the fewest possible components, and it successfully did so by surprising the humans with a single wire, i.e. an antenna picking up nearby stray RF.
That is my favorite part of GA. Gradient free optimization but it turns out making a good fitness function is hard and like 70% of the time it just exploits some assumptions or gap you have in your theories. Really reveals the problem in different ways that traditional ML.
As someone who does a lot of genetic programming (like, old-school, without AI/LLMs, etc), I can confirm that the fitness function is very difficult to get right, especially if you are trying to evolve programs that have "adversarial fitness" -- you'd need to maintain a hall-of-fame, and that just makes the runs take _much_ longer, because, chances are, your fitness function is the bottle-neck.
So, it is very hands-off, but also very expensive, and it is never clear if optimizing the fitness function is worth it, because the fitness function itself may be insufficiently or incorrectly specified.
However, I do think that people should try, even with just a whiteboard or a notebook, to design a fitness-function, for their problem, as if they were going to try to evolve it, because (1) it forces them to explicate their correctness constraints, and (2) they may discover that the program that they are trying to write _is equivalent_ to the fitness function.
I'll give you an example for point 2. Many years ago, I had to parse a gnarly language, and I chose to do it via Chomsky Grammars (that automatically build a tree based on the grammar-spec). Chomsky Grammars are cool, in that they are basically just a state-machine, but they are incredibly difficult to debug: when they work, they might work incorrectly (malformed tree), and when they fail, they give no reason for failure (because even with a trace, you are trying to figure out which backtrack should not have happened). So, out of desperation, I started to consider using genetic programming to just evolve a correct Chomsky Grammar. It became clear that there are only 2 possible fitness functions (1) a function that tests a hand-picked input against a hand-crafted tree-output (which is vulnerable to over-fitting), and (2) a function that is not (well, is much less) vulnerable to over-fitting, but is effectively a pre-existing, correct grammar that can produce those trees.
If you are in situation 2, then the genetic programming is not necessary, unless you are trying to create an optimized (or obfuscated) parser, and even then the optimization may be overfit to the test-inputs (even if they are generated test-inputs from the grammar itself). If you are in situation 1, then you are better off re-evaluating your approach (I abandoned the Chomsky Grammar notation, and invented one that is much easier to understand and debug, without losing any of the expressiveness -- it also happens to be slower, but fast-and-broken is worthless compared to not-so-fast-and-works-fine).
One place where genetic programming has been consistently awesome, is in parameter-search style problems (e.g. your genome is a long list of floats, representing weights and/or anti-weights, and you need to find out which weights give you more fitness (or less error)). I hear good things about variable-neighborhood-search, but have yet to try it.
That sounds apocryphal but there was a noted paper describing a frequency discriminator implemented using a genetic algorithm and it ended up tied to the exact piece of silicon used to evolve it, with logic cells not connected to anything still changing the output.
https://osmarks.net/assets/misc/evolved-circuit.pdf
This too: https://en.wikipedia.org/wiki/Evolvable_hardware
Starting with: https://sci-hub.ru/storage/moscow/4324/11d145b2c2c3ab320f70b...
That second paper is absolutely amazing, I’ve always heard this story and never bothered to find the source.
The section with oscilloscope traces showing the progression of the “designs” over time was extremely interesting - I’d love to see what the 10x10 grid of functions looked like at each snapshot.
Thank you!
The other side is Cognitive Radio [1] which also evolve the OTA protocols for cooperative diversity from IEEE 802.22 onwards. Now I can see AI, via a local SLM/NPU plus agentic GNURadio loops for new radio use cases. This is going to be much more wide spread in the upcoming 3GPP 6G releases in 2030.
[1] https://en.wikipedia.org/wiki/Cognitive_radio
GA’s optimize only combinatorial problems though — where you have discrete set of choices (~genes) for each variable, and therefore do not have a gradient
You can use a GA with continuous parameters and a smooth gradient but it probably isn't the most efficient method in that case.
One great application of AI design is patent poisoning. Use AI to churn out masses of variant designs, make them publicly visible on a web site, and if future patents come out use any collisions to invalidate them or at least restrict their scope (generalization of a patent is limited by prior art.)
I’m reminded of lawyer Damien Riehl’s (performative) reaction to the Sam Smith infringement decision, back in 2019/2020. He and programmer Noah Rubin algorithmically generated every possible melody (within a certain combinatorial space, in MIDI format as I recall), and purported to release them under CC-0 license [0]. He went on to attract some attention and explain his argument at a regional TEDx event [1].
I seem to recall legal commentators reacting with an eyeroll—apparently judges split much finer hairs than these for a living—but it was a cute stunt.
[0] https://allthemusic.info/
[1] https://m.youtube.com/watch?v=sJtm0MoOgiU and https://www.the-independent.com/tech/music-copyright-algorit...
Copyrights are a bit different from patents.
For sure, but I suspect the law might look similarly dimly on the argument that “machines systematically generated all possibilities in the problem space” === “somebody already had this idea.” I’d imagine maybe by reading specific human intention into “prior art” and “existing work” and those sorts of terms.
Which is not to say let’s not do it anyway and see!
Or they might conclude "even an AI could come up with this, it's obvious". Obviousness is a defense in patent law, not in copyright law.
There is a fairly effective counterpoint here that information is worthless if nobody can find it. Generating a handful of useful designs in a sea of pointless ones doesn’t count as ‘coming up with it’. Remember that a description of every idea already exists in the Library of Babel. I do not think that AI generation without curation really counts as ideation.
If patents can’t be granted to an AI inventor, I don’t see how such AI “inventions” could be used as prior art.
Prior art doesn't have to be a patent. A frickin' Soviet movie was prior art in the Blue Origin/SpaceX lawsuit about landing rockets on barges.
Yes, but that movie was made by a human.
My point was that it’s hard to imagine citing something that could not be patented as prior art. It would be like citing a phone book as proof that a software program can’t be copyrighted (“the exact bytes appear in the 1973 Albany NY white pages, therefore it wasn’t original”)
Wouldn't work. Judges would not treat the AI generated designs as prior art without proof of human involvement (above and beyond entering the prompt).
Maybe not prior art but very much "not novel".
"Humans couldn't even imagine" seems like overselling it, but I'm sure that machine learning algorithms can brute force their way to chip designs no one has tried before and that some of those might be useful to us. That seems like a pretty reasonable thing for a computer to do.
Here's a story about a genetic algorithm evolving a circuit that works with logic gates that aren't even connected, seemingly by using magnetic flux: https://www.damninteresting.com/on-the-origin-of-circuits/
Thanks, I was trying to recall that article. Fascinating stuff to this non-expert.
It's marketing bullshit. For one, it's like proving a negative; you can't prove to me that humans couldn't have imagined it. Second, humans have already imagined quite a lot of crazy stuff...
It really just means, irregular, unconventional, not in line with traditional designs.
Machine learning layer cake with some brute force crumbs.
I wonder if our common expectation that true theories somehow had to be beautiful and elegant is going to survive the coming century. What if "real" nature phenomenon were actually best described by horrible mess of impossible equations, that only machines could actually manipulate and reason about ?
That would be really sad..
> our common expectation
I think you're going too far with this. Most people understand scientific theories to be an approximation. F=ma is approximately true, in the sense that it's only accurate within the newtonian regime and each of those terms includes so many asterisks that you will only ever measure it approximately.
The latter is the jokes about the physicists "assuming a perfectly spherical cow."
In fact that's kinda the whole point of the "unreasonable effectiveness of mathematics" essay. It is unreasonable that mathematical approximations are so good at describing our world.
> The latter is the jokes about the physicists "assuming a perfectly spherical cow."
Not to detract from your point at all, but I only ever heard this joke about mathematicians!
I've definitely heard a similar joke about economists. It probably applies to most sciences, tbh
I often think this about medicine and the human body. We want to believe that our bodies are some miraculous well oiled machine. But it often seems that it’s a barely held together bag of mess.
I think politics and economics work along similar lines.
Biology is incredibly robust!! I'd say barely held together bag of mess describes something like an internal combustion engine. A primate, on the other hand, is a self-replicating machine capable of self-repair and just about universal fuel sourcing. It has a robust defense network capable of identifying and eradicating a staggering number of foreign replicators. It has holographic design storage, with each cell containing the plan for the whole organisim. It has general cognition based on a world model, and does all this on almost no energy.
Biology is incredibly well oiled!
I think your reply and the parent can both be true, you're just using slightly the same words to describe different things.
The parent is talking more about elegant simplicity vs. sprawling, seemingly haphazard complexity, and you're talking more about durability to failure points and 'completeness'.
Likewise, in code, a lot of the most durable, battle tested software looks extremely inelegant and duct taped, as 90% of the code is dedicated to handling one-off patches and weird edge cases.
No, you just need to be on a regular good maintenance schedule!
My suspicion is that we had a sense that generality and compactness was really neat, so we liked easily-remembered laws like F=ma. Applies everywhere, is clean.
When you attempt to hyper-optimize, even with humans in the loop, you end up a mess. You're lucky if you can find clean guiding principles anywhere. If you can hyper-optimize hyper quickly, you end up with an extra layer of mess.
Or that tractable models are just more useful than intractable ones even when the latter are more accurate.
This has been on my mind lately! Especially in light of the many incomprehensible but machine-checkable proofs we've been hearing about.
Occam's Razor is a useful heuristic, but it biases us towards simpler explanations.
But those proofs are showing that the fundamental axioms (which are generally simple and elegant) are still enough to build a complex result.
I think of elegance as not having to add epicycles, not that everything in the system has to be simple.
Also, without a working theory the, the space of possible solutions is near infinite. LLMs manage to pluck out the space of comprehensible English strings from n-dimensional hell. Even if this is done with a black box of billions of parameters, it’s still elegance in the sense that such a space even exists and was found
Would it be sad? If it’s gnarly and it solves the problem, as an end user I don’t really care. The only people who lose are the mathematical purists
Math is a language to explain systems. Teaching someone that force varies linearly to mass is a helpful first pass. It isn’t exactly linear but is not exponential at all.
Gaining expertise is always the hard part and our new LLM overlords are making that much harder. So the simple “pure” functions as a teaching aid have never been more important.
End users have never cared about how the sausage is made though.
> LLM overlords are making that much harder.
LLMs can explain complex things to humans with tons of specific context that you don’t find in textbooks or even a google search.
It’s probably never been easier to grasp a large codebase than it is today for example. You can probe and ask specific questions without going through a maze of imports and relationships and config files yourself.
Learning things will always be up to the person, it’s still a choice and dedication to a craft can still be taught.
> LLMs can explain complex things to humans
I keep meeting people who think this and have enormous understanding gaps in the topics they've had an LLM teach them.
The absolute worst judge of how well someone understands a complex topic is the novice themselves.
The difference now is that the learning is optional (more often but not always) to getting the task done.
When gaining mastery is not a requirement to doing novice-level work, many fewer people will get there. It takes more dedication than it did before.
The "common expectation" I think, misses the point. The idea isn't that fundamental theories are simple or elegant (quantum physics equations are pretty darn ugly), it's that, given the choice between a more complicated and a more simple theory, generally the simplest one is the most accurate choice.
I don’t agree with that at all. Maybe for asinine things like human behavior but otherwise nature and physics don’t really follow that rubric.
Are you thinking of any specific examples? I don't disagree that complex things generally end up having complex explanations, but I'm admittedly drawing a blank trying to come up with things where the most complex explanation ended up being the correct one.
There are many marvels of evolution's ability to come up with robust complex distributed systems which work way better than anything we build. The one I've learned about most recently is the https://en.wikipedia.org/wiki/Immunological_synapse in which different kinds of white blood cells gather around a bit of evidence that one of them found and decide whether to shoot the messenger (clonal anergy), or raise a clone army to defeat the invader (T-cell activation).
Imagine that it's maybe the 1800's and you're asking why somebody who has already survived smallpox is not susceptible to becoming infected again. If you offered an explanation involving tiny detectives wandering around and collecting evidence which they present to each other and decide whether to multiply... one in which the tolerance comes from the detectives from the previous fight still hanging around in your lymph nodes ready to spring into action if they run across the right kind of evidence. Well that would probably be a more complicated explanation that anybody at the time would offer, and it would also be correct.
Incomplete prior knowledge doesn't mean it's simpler, just that it's inaccurate. Would the phenomenon you're describing really accurately be explained by something _simpler_?
You should look into Solomonoff induction. Nature and physics, absolutely, tautologically, have to follow the "shortest explanation is more likely principle".
https://en.wikipedia.org/wiki/Solomonoff%27s_theory_of_induc...
All theories are wrong.
Some are useful.
Having theories that only give answers, but you can't reason about is not as useful. Having a theory where you don't know the limits of it's applicability, can be very dangerous.
At least in the physical realm there is not yet anything that combines relativity with QM so they can only be approximations. Even in math so far there seem to be similar challenges using programatic and "AI" driven solutions and proofs.
Still, I know that LLMs will be useful for Verilog/VHDL and particularly with verification, where they are already heavily used. Defined outputs and complete test coverage is already such a big part digital/asic design, I'd be surprised if it isn't used a lot more. Many software people would say that hardware is badly written copy-pasta, as it is. That said, higher velocity slop and hardware "technical debt" isn't something you can fix with an update. And no matter how fast you "ship", you won't get parts back in less than a few months. Poorly used, it will lead to expensive failures.
That is very unlikely due to Solomonoff induction...
Solomonoff induction doesn’t concern itself with what is truth and reality. It just says which theory to prefer and how to determine so objectively when multiple are equally precise in making predictions of observations. It’s a formal description of Occam’s razor.
OPs argument is that reality is expressed by very complex equations and interactions; by definition this is outside of Solomonoff induction because it’s easy to imagine this accurate model by definition is the shortest algorithmic explanation, it’s just orders of magnitude more complex than our current approximations.
You should include the error correction code length in the description length. This means Newtonian mechanics was a much longer theory to describe Mercury's orbit than general relativity. It was only the shorter theory before they had the data showing a discrepancy. Which is the correct approach to describing your reality, because until you see a discrepancy, the extensional properties all follow the shorter rules.
I guess the argument from OP would look like: "Yes, now imagine we poke and extend our universe as far as we can. How much bigger do you think our final 'shortest description' would be? I imagine it may be orders of magnitude more complex."
Well, I can imagine a squared circle... doesn't mean the math checks out. I would reply that you do not have to imagine, you can go about looking at different mathematically possible universes in Tegmark IV and find the expected number of bits for the one you actually exist in. Which is ~0 bits more complex than the shortest description based on the data you currently have.
Also, note that Newtonian mechanics is not actually a very short theory for building a universe, because you have to instantiate every object in the universe. You actually get a lot more of the structure for free with general relativity (re: Wigner's classification of the particles). An observer in a presumed-Newtonian universe calling it a simple theory would be like saying, "I compressed Wikipedia to one byte, just by putting it all in the decompiler!"
>I wonder if our common expectation that true theories somehow had to be beautiful and elegant is going to survive the coming century.
That's the layman's idea of physics theories. They are beautiful and elegant only on the surface, that's why they're technically models and approximations of the real world. The standard model renormalization techniques are a mess of patches and ad-hoc heuristics, pretty far from the "this lagrangian literally contains all physics". Generally you just _ignore_ higher order terms and just call it a day. The famous E=mc^2 it's just the first term of a Taylor expansion. The beautiful form of physics it's what you would call "good enough" and often just a pedagogical tool.
> The famous E=mc^2 it's just the first term of a Taylor expansion.
Is this actually true? My understanding was that E=mc^2 is exact for a particle at rest.
It would represent a pretty sharp inversion from all the progress of mathematical physics until the present.
Up until the present it has been a nearly uniform march of revealed symmetries, collapsed privileged frames of reference, and other such (in the deepest sense) simplifications in our model of reality that has improved its fidelity to the measurable.
I hang qualifier about these developments being simplifying because the result isn't simple in the details: quantum chromodynamics is a daunting subject! But it's not just an enumeration of details and contradictions, the particle zoo that preceded the Eightfold Way looked like line noise, now in indexed notation the Lagrangian of the entire Standard Model fits on a page (or so I've been told I've never actually seen the page).
It's almost tautological that the frontier where it's still messy involves an unrevealed symmetry or a persistent privileged frame of reference, that's what frontier means, we don't see past it to the seam where it folds up.
Personally I suspect AI systems will be a great deal more inclined to discard the parochial axioms that have every point placed human ego above simplicity.
It doesn't resolve all of the open problems in physics if you amputate consciousness, free will, agency persistent identity, and an unambiguous arrow of time.
But it starts looking possible to make progress.
Hallucination, when it succeeds, is the intelligence.
It would be really cool. We already know everything at the lowest levels is a probability cloud. There’s beauty and contentment in not really being able to nail anything down for eternity…
That's a result of the Copenhagen Interpretation. There are other interpretations of the math which don't rely on reality fundamentally being a probability cloud/wave/field.
I’m not well versed in this but if fundamental particles are probability clouds, the future is not deterministic.
Why not? Every cloud that matters is resolved. What's to say a different resolution to any cloud can be possible?
Does the nature look like a horrible mess to you?
If you’re not being facetious, then the answer is very much a yes.
How would nature be best described as a horrible mess of impossible equations? They would be best described as elegant and beautiful no?
I think your point is more that we might be able to initially describe complex phenomena as messy, horrible complex equations, that doesn’t mean we shouldn’t work to simplify them and make them more understandable to us.
Look up diagrams for cell signaling cascades sometime. It's emblematic.
In some ways it's like a smalltalk system haha. I think you can find some elegance there. The system is overall complex, but with simple primitives.
I’m a bit frustrated. AI can do a looot of things; but I think as we continue to muddy the waters between LLMs and more traditional machine learning like Monte Carlo, Genetic Algoriths, Expert Systems and other Statistics magic tricks, we’re too aggressively conflating established and morally neutral activities in ML with the concerns that people have about LLMs and Stable Diffusion.
Though I also imagine that that is the point.
It is a problem because people will talk about what AI can do implying that an LLM can do that thing, making it seem like a pure LLM can do almost anything. On the other hand people will say AI will never be able to do X because an LLM can’t do that thing well natively. AI has become too vague of a term to be useful.
We're relearning that intelligence is spikey, and that different things that we consider 'intelligent' can have vastly different capabilities.
We're learning that people are way too lax with where they apply the term "intelligent". LLMs aren't remotely intelligent, but people are trying to ride the hype train and call them intelligence.
> LLMs aren't remotely intelligent
Maybe I'm just significantly and unrepresentatively unlucky, but Claude is significantly more intelligent than the average human around me on most any metric I can think of.
Very much indeed. The term itself is not properly defined, strictly speaking.
this is just false.
by any meaningful measure of intelligence. the latest models are much smarter than the bulk of the population.
how would you define intelligence?
“Intelligence is the ability to learn.”
That is a meaningful measure of intelligence that every LLM completely fails at.
I wish I could wave a magic wand and just make the word "AI" go away. It has no actual meaning. It could mean anything from your opponent in Mario Kart to Stable Diffusion.
"AI" == "what (through tech) can replace a professional"
It may seem similarly vague, but it does in fact open interesting, productive, and necessary questions. A "computer" was a professional crunching numbers - "replaced", "easily" because of the deterministic procedural nature of said work, but what about the technical effort to arrive there, and what about the less "mechanical" jobs? When do "processes" become "intelligence"?
Some of us had studied AI originally to study the mind - "how do we formalize thought". It's the interdisciplinary, transversal nature of the area.
Also maybe compare that with that large and important intersection between CS and Economics - the "science of optimization" and its implementation in efficient IT systems. The effort in terms of that different discipline may not be evident, yet lots of engineering is "optimizing" and the generalization of those solutions we call Economics (see the book Algorithms to live by).
So: the term "Artificial Intelligence" may not be important as CS solutions to practical problems are built (you just focus on the better solution), but there is relevance to the "side disciplince" of AI, and from that perspective that is the cone, the scope anyway. "How would an intelligent solver approach the problem".
> "AI" == "what (through tech) can replace a professional"
But as you point out, we used to have human calculators. So is a simple desk calculator a form of "AI"? If so, what type of software isn't AI?
> is a simple desk calculator a form of "AI"
If what it does is "taking care of the carry", it represents a pretty minimal requirement for intelligence - it does replace a professional that could do it, but that professional does not have to apply too much proficiency and cleverness to do its job. It is improper AI.
> what type of software isn't AI
That which would not correspond to the job of an intelligent entity. Maybe blitting bitmaps around a screen?
As I tried to convey, it is more of a matter of perspective: the area of "implementing ways to solve problems as an intelligent entity would". It is a discipline that intersects others - engineering, logic, brain science, philosophy, epistemology, maybe again economics (as "the science of optimality and efficiency" - as an intelligent solver would do)... Consider it a special discipline that spans many other realms.
> Maybe blitting bitmaps around a screen?
Okay, that makes sense. Even so:
> If what it does is "taking care of the carry", it represents a pretty minimal requirement for intelligence - it does replace a professional that could do it, but that professional does not have to apply too much proficiency and cleverness to do its job. It is improper AI.
I think you're underselling how much mental work is required to solve complex arithmetic. Yes, it's simple for a computer, but (1) even basic computers are extremely complex in absolute terms, and (2) even the most complex computing tasks could be considered simple once you break them down far enough—for example, a large language model is "just" fancy matrix multiplication.
So I feel like there's a "sufficiently advanced technology is indistinguishable from magic" element here. Something becomes AI once it seems sufficiently advanced. But then time passes and it doesn't feel that advanced anymore.
I understand that human language doesn't always have a super precise definition, and I'm not trying to be pedantic. I think the term "artificial intelligence" is under-specified to the point of having virtually no meaning. To the extent that it is useful—obviously, a lot of people are using it conversation, so something is getting communicated—it's because it's possible to infer from context what someone is referring to (ie "the student used AI to write her essay" is clearly referring to an LLM, not Eliza).
We'd all be better off if we used words that describe what we're actually talking about.
Game AI uses behaviour trees, usually coded by hand. Decision trees are used for classification and are normally learned from data. The latter are a traditional AI technique from the early days of the modern machine learning era, in the 1990's.
I disagree. AI is doing exactly what it was predicted it would in science fiction.
The computer can now literally talk to you in natural language and then perfectly produce sophisticated actions in response to completely arbitrary and unstructured input. It trivially passes the Turing test. By any definition prior to the year 2023 we are living with Artificial General Intelligence and it’s here now.
Current LLMs don't pass the turing test.
Remember, the interrogator is allowed to be hostile, so they would obviously employ all known prompt injections and typical LLM 'gotchas' to figure out who the AI is.
So where are the androids? If it's AGI, why is it used as a tool, waiting to be prompted or executed by humans? Where is Skynet? Military applications still rely on human operators.
Robotics is advancing a bit slower, but is making progress as well.
Yes, but unlike a lot of science fiction, robots, LLMs and other AI remain tools for human use. Augmented Intelligence would have been the more accurate word for real world AI.
You realize llms as a field is barely 5 years old? Give it at least another 5.
I doubt LLMs will give us full embodied intelligence that science fiction androids have. Maybe some other approach. But I suspect for the forseable future LLMs, robots and other AI methods will remain tools, not independent agents like Star Trek Data or Skynet.
VLAs are new LLMs. Give them 5 years to develop. But even good old LLMs are still improving every six months.
Yeah. We'll be arguing "is it really AGI" for many more years. Meanwhile, everyone interesting is going to have moved on from that question, choosing to spend time on "who cares if it's AGI, can it do $foo", for whatever value of $foo is interesting to them. Whether the machine is folding clothes or folding proteins, AGI isn't well defined other than "I'll know it when I see it", so whether or not it's AGI, the question is what job is the machine capable of and is it cheaper than a human? A humanoid robot that can work a warehouse is not putting anyone out of a job if it costs a billion dollars, and neither is a digital AI employee that costs the company a billion dollars either.
"AI" is a term cursed by cool sci-fi implications. It makes it a kick ass marketing term because most people are going to have some familiarity with sci-fi AI and "X media predicted Y technology" is a pretty widespread belief for a lot of values of X (star trek, Hitchhikers Guide to the Galaxy, Arthur C. Clarke) and Y (internet, cell phones, VR). If you want to tell someone we're making big strides in something, linking it into some popsci understanding of sci-fi being the great predictor of human achievement is low effort and high impact for quite a few people.
People aren't trying to communicate accurately if their first priority is getting you excited about the thing!
I miss "predictive analytics". Too boring and honest for marketers though.
I have been practicing saying ML for traditional machine learning and LLMs for LLMs for just this reason. Trying not to say AI anymore. Too ambiguous. Sometimes I'm talking about game AI even, I'll try to use shorthand for whatever algorithm I think the AI is using (often I'll talk about its flowchart, though not always sure it's literally using that under the hood).
What is ChatGPT then? Sure it's an LLM, but I can give the app pictures and audio, and it can generate pictures for me. Do we distinguish between the bits of the architecture to accomplish those features separately from the LLM part of the product?
Yes? Or just call it a chatbot if you don't care about the implementation details.
Recently I heard some people conflate procedural generation and generative AI and I had to explain why there isn't some legal or ethical issue with what breaks down to essentially scattering some points.
It's really getting annoying having to have these conversations.
Just as more successful machine learning fields distanced themselves from the term during the AI winter, I suppose we will (and perhaps are?) be seeing them adopt it again, now that we are in an "AI summer".
As always, it's a matter of funding. Both inside academia and outside of it. I remember when nanotechnology was all the rage. Everyone flocked to writing grant proposals about their "nano" technology that was thousands or millons of nanometers, aka micrometers or even millimeters. Stupid but if it works it's not stupid. The old joke is what do you call AI that works? Machine Learning.
The real question is how much compute do you need. With LLMs getting popular, so is compute. That's the real win for non-LLM technologies. The sheer availability of GPU capacity. Yes, it's expensive, but time in a GB300 supercomputer isn't even possible if they don't exist.
Alexnet succeeded for many reasons but a big reason is that computers got good enough to apply those algorithms and techniques in practice. Outside of LLMs, what new AI/ML systems await us in the future? The LLM bubble popping, if it ever does, is going to leave us with supercomputer capacity going unused and available for cheap, meaning experiments that were once infeasibly expensive become practical. I can't afford $10 million to run a weather simulation, but at $1,000 for the same amount of compute, a lot more experimentation becomes practical.
Reinforcement learning can solve a Rubik’s Cube. A LLM that hasn’t been trained to solve a Rubik’s Cube can not.
> AI can do a looot of things
AI is not a real thing or a natural kind but a perspective. Whether something qualifies as "AI" or not cannot be decided by the objective features of the thing. Ergo, it can be defined at the author's pleasure.
> conflating established and morally neutral activities in ML
LLMs are no more or less morally neutral than other ML techniques.
[flagged]
This is such a strange attitude to have. How can you look at the history of the world and be so caustic to concern?
strange how? in the history of the world, had there been a technology that's been wished out of existence by the concerned?
Leaded gasoline.
you know that's not what I meant. also, avgas.
What did you mean, then?
Also, CFCs.
things like the ones I had enumerated in my first comment. I'm old enough to remember FUD about all of those and many others, and perceptive enough to identify many ongoing public opinion campaigns as such.
Phrenology
the biggest question for me is how robust are these designs.
in the journal articles they did show measurements of real devices which agreed fine with predictions, but i didn't find them addressing it explicitly in the text. also, some systems they presented contained subblocks that were conventionally designed that could be carrying some of the weight.
or maybe i'm just sour that they're coming for my job? or maybe that's what they want us to think?
i think what wins in practice is simple ideas that can work in spite of all manufacturing and environment variations, and model limitations -- think stuff like feedback and symmetry. and what they show here is the opposite of that. i've done blind optimization of circuit parameters some times only to end up realizing some pretty simple such ideas that i'd missed (like "you need symmetry here" or "you just need more bandwidth here") and made complete sense when you thought about them. so i wonder if we can't tweak a few pixels in their structures and reveal something simpler.
also, obligatory mention: "genetic antennas"
They address this in the conclusion
> How generalizable are these methods? Can they consistently deliver truly high performance? Can we get to a place where AI produces designs that maximize every conceivable trade-off, holistically optimizing every parameter to its most ideal physical state? .... AI can hallucinate a design that creates bad circuits that don’t work. This means verification methods need to remain under human oversight.
And they are essentially correct. We need better validation and verification methods, both software and hardware to keep in check the mistakes of automated random processes.
> but i didn't find them addressing it explicitly in the text
Yes, this is exactly what bothers me about this article and about a few similar articles published in the past, that they do not contain any evidence that their claims about the usefulness of AI in design are true.
In TFA it says that the role of AI is replacing the electromagnetic simulator in the optimization process, by guessing the behavior of the structure, which is many orders of magnitude faster than a simulation.
This sounds plausible, but in order to believe this I would want to see the differences between AI guesses and real measurements, in the case of structures with geometries that are very different from those used in the training of the AI.
Also I would want to see exactly with which simulators they have compared the speed of the AI model.
There are various simulation approaches for electromagnetic fields and electronic circuits, that can trade-off accuracy for speed, so I am not convinced that AI inference takes necessarily much less time than some faster low-accuracy methods of simulation, which would still be more accurate and more reliable than AI guesses.
I came to mention genetic antennae as well!
Since you beat me to it, I'll add something that relates relates you were saying on "realizing some pretty simple... ideas".
I think a big plus of computer aided design like this is "innovization"[1]. Somewhat awkward term. But, a system like this leading one to deeper understanding of a particular process is the general idea. It's a fun feeling in practice.
[1]: https://dl.acm.org/doi/10.1145/1143997.1144266
Yeah it's a hype slop piece
> the biggest question for me is how robust are these designs.
Maybe it doesn't matter?
I mean, of course it matters. But most of this sort of design space is effectively NP-complete, where the creation starts with a blank schematic page and has an impossibly large search space, but where the checking of the design is much simpler.
> also, obligatory mention: "genetic antennas"
Exactly. How does this work? When confronted with the question, of course, everybody gets all excited about the constrained randomness of the GA, but if you think about it, what really makes it work is that there is a comparatively cheap test for fitness for purpose.
Reminds me of this old article - https://www.damninteresting.com/on-the-origin-of-circuits/
One of my favorite little morsels of internet goodness.
I was going to post this as well, its delightful to see that other people enjoy it since it was really mind-blowing when I read it.
It's interesting since I saw another comment near yours that raised the question of robustness of the lab-grown design, which I thought was kind of the most fascinating part of the damninteresting article was the revelation that the evolved programs were inseparable from the single physical FPGA used in the training. Since this RFIC training model employs a simulator, do you suppose that the quirks of the physical hardware on which the simulator runs are sufficiently isolated from the training such that a pair of designs would behave similarly when the simulator was run on distinct hardware? And I guess the even more obvious question is whether a design evolved on a simulator would have any hope of behaving as expected in physical hardware?
My hunch about the latter is no, although it still seems like an interesting study, and I often find myself thinking that really understanding what was going on with the FPGAs might be a prerequisite for really understanding how to master reinforcement learning.
Anyway I'm glad you posted this and if you have any other favorites related to this domain send them my way!
Yes. An example of a species so specialized and optimized that it can no longer adapt. Also, an example of POSIWID.
One takeaway from the article is that they had to get rid of the tried and tested fundamental building blocks of chip design to generate this advancement. I wonder if the same applies for mundane coding. Are the incredible innovations in AI coding actually hampered by rust and python? Should we let AI tools just code in the lowest level possible?
I’m intrigued by the question but I do think it has some worrying implications for portability.
No. I'm really not fan of Python but you're not going to parse JSON at assembler level. You need to choose right level of indirection for each task. And LLMs are good at more than Python and Rust. I'm using LLMs to write programs in my own language that is compiled by a very fringe language and it happily does everything from scripting on ESP32 to audio plugins and 3D gfx on desktop.
It's not really that magical. As TFA points out, RFIC design, way beyond normal RF engineering, is close to black magic that relies a lot on the knowledge and experience of the designer, assisted by what would have been supercomputer-level-a-few-decades-ago modelling and design tools. What AI can do is a breadth-first exploration of all possible outcomes and then pick the best-performing one rather than the human-level "this seems like a good path to go down, let's explore it further".
Does it need to be magical to be interesting or useful?
> But freed from the constraints of human-designed templates and the need for humans to even understand the rationale of electromagnetic structures, ... ICs ... can take on truly wild-looking yet efficient designs.
I feel like technology is going to become alien at some point. We're all going to be using magical runes instead of chips.
I work in a related field and “inverse” design is what this is called. Such designs usually are not manufacturable. I’m not too worried about my iob.
That said we’ve had some success internally having Claude do parameter sweeps
> That’s not even to speak of all the movie plots that would have been ruined.
I clicked on all the links. Pretty much all of those movies could still work with wired technology. Even the one called cellular, in which a woman is trapped in an attic with a broken landline phone and manages to connect wires and dial a random number.
Yes I'm nitpicking. I guess I'm glad we have Wi-Fi and all, but don't try to sell me on it as a crucial plot device
And what about all the movie plots ruined by the existence of cell phones?
People seem to (in)conveniently have no service in movies a lot, exactly when they need it
Very interesting. I wonder how hardware craft would look like after adapting AI in a massive scale.
In case anyone feels déjà vu, Popular Mechanics wrote about this professor's lab in Jan 2025, with almost the same title: "AI Designed Computer Chips That the Human Mind Can't Understand".
I feel a bit of unease when I read this title, not because of the threat of AI, but because the prevailing aphorism that "RF is black magic" is a slap in the face to the millions of physicists and RF engineers who DO understand every bit of this. It's a fun harmless anti-intellectual saw that I don't believe is harmless at all. We need more RF engineers and telling people it's all "black magic" and "wizardry" (and worst of all, saying "even RF engineers don't understand RF") makes it seem like it's not worth studying.
> telling people it's all "black magic" and "wizardry" (and worst of all, saying "even RF engineers don't understand RF") makes it seem like it's not worth studying.
I think the opposite is true. It being advertised as difficult to understand is one of the reasons I personally decided to study RF Engineering. The prospect of learning something so challenging pulled me in. The Smith Chart helped.
> Freed from intelligibility and aesthetics, AI designs faster
I like this headline. In other words, AI will suck out every last bit that makes engineering fun.
I know, I know. The job is to make money for your employer not have fun. AI makes money faster so shut up and do your job.
But fuck, I took this career because I found joy in understanding things and making things that look and work well.
Hopefully one day AI will design away the need for popups and other-things-that-prevent-you-from-reading-the-damn-article.
Yep, can't wait until everything is free and costs nothing to generate content. Free hosting and electricity will be super sweet too. Won't need admins or even the Internet. Everything I want will just be free for me because I don't think anything has value.
[dead]
Ai can’t even place and route a two layer board with a microcontroller and a few peripherals.
Great, and here I thought my job was safe.
The comments here are trending towards "There's nothing new here, I could design 5g radio chips with a cheap linux box running FTP".
What a well written article
The methods outlined in this article aren't new. Scientists were using "genetic algorithms" to design antennas that weren't understood by anyone, but worked well, decades ago.
Chips? I've tried to task Opus, Gemini and Codex with a simple PCB. All of them placed holes correctly but can't understand that the traces should not cross physically.
The AI in the article isn't an LLM.
Read the article.
I did my PhD on inverse design of electromagnetic structures. I really hate that we're calling this AI when there isn't any training, really.
>To arrive at the “sweet spot” where all these different parameters are balanced into optimal harmony, designers will typically lay out several different versions of the circuit, using intuitions and methods they have picked up in their years of training.
I kind of thought the real success is when the designer comes up with key things that are well beyond their training or any training that could have been done up until that time. Based on their years of experience living in an environment where training is table stakes but that's not the thing that's relied upon the most in the end.
With LLMs it seems like odds are that a concept which is statistically insignificant in the training set may surface in place of a truly novel solution, effectively displacing the real breakthroughs that actually go beyond trainable performance.
In a way that decision-makers can not tell the difference, and that could be the worst part.
We have always known the old trick of genetic algorithms to produce better radio chips.
The problem isn’t the design: its manufacturing restraints.
This is nothing new or impressive.
Then why can't these constraints be encoded into the selection/scoring function ?
Because you might actually want to manufacture one offs, like for space equipment.
Now let's get them to come up with a valid design including a valid QR code. Maybe one containing Maxwell's equations.
I don’t know. I can imagine quite a bit.
Unexpected Star Wars! A surprise to be sure, but a welcome one :)
If you don't know how it works, then you don't know that it works.
How does your consciousness work?
Pretty well, between all the hallucinogens
Utterly haphazardly and inconsistently of course, same as yours. You thought that was some sort of argument? It slots right in and contradicts nothing.
[dead]
But is this AGI?
[flagged]
I am confused, every day I read on HN that AI's can just interpolate the data they have seen in training, and that they are structurally incapable of coming up with something new, creative and not in the training distribution.
This is analogies to finding a new prime number by brute force using existing maths, rather than inventing new maths to get there.
The AI in this case didn't create a novel technology- it merely used the existing technology without basing the new design on a previous one. The whole "human couldn't come up with it" is because the possible design space is so large, there's no reason a human would start where the AI did.
The thing the AI did better than humans was brute forcing a solution faster. Still a very handy thing to have, but it isn't "creating" in the sense that it invented new materials or fabrication processes or anything novel.
> I read on HN that AI's can just interpolate the data they have seen in training
No. That can be said about LLMs, but not about all forms of AI. The technique used is not a LLM.
Sadly we've bastardized the term AI that, if it ever meant anything, it's meaningless now. The currently most voted thread in this post discuses the topic.
Have you read the article? The creative element came from the researchers:
> In our new approach, the architecture begins essentially from nothing and is progressively assembled through successive iterations. The system explores the design space by generating myriad candidate circuit combinations and mapping the resulting performance trade-offs as it navigates this landscape. Because the process is not biased by prior human design choices, it can produce completely novel circuit topologies that look markedly different from those created by human designers.
In my experience, if you tell them to research the web to see if their idea has been pursued before, you can get them to keep proposing new things until something is sufficiently new, even if it's a new interpolation between existing concepts, that it's effectively an original idea.
This is wrong - the training data is necessary but insufficient. There are a lot of other parts of the architectures used that add a lot of value - otherwise Markov chains would be all you need. There are layers upon layers with non linear activation functions, learned residuals, etc. They still absolutely must interpolate but the space they interpolate through is much more complex than the training data, and they can definitely create things not in their training data. What they can not do is wander outside their non linear parameter space’s convex hull. But this is a really permissive constraint on what they can do “creatively.” People generally under estimate the advantage the architectures confer on that constraint. This is why there was a step function change in expressive power as the architectures (attention, self attention, transformers, diffusions, others) evolved given the same training data. Generally though I challenge you to define “creative” in a way that is precise enough to measure and isn’t self referential or refer to concepts ill defined.
The key tho is can they solve problems not easily solved before with prior techniques. Further can they identify problems not readily presented. Then identify novel solutions. Etc. The answer is emphatically yes they can. These features don’t have to literally exist in their training data, but the supporting highly convoluted network of associations of all their training data does have to in some complex space allow for it to produce these answers. It’s not the same as they’re stochastic parrots at all.
Are they creative? No, because they don’t have awareness. My personal imprecise definition of creative requires both self and awareness as well as free will. There is no driving awareness in all AI architectures, it all derives from extrinsic impetus. Creativity is derived, IMO, from a layer of our minds that is not readily assessed or measured and is only indirectly expressed through language, art, and music. Hence it is not directly trainable and therefore a learning model can’t learn it by reinforcement. It can learn the proxies, but the proxies are not, as we all deeply know, the same as our experienced awareness. We are not our words, our art, our music. We try hard to bridge it, but it’s impossible and you and I know this to be true from experience. In fact we can not even examine our own awareness because it’s not directly observable or possible for us to directly reason about. This is core to a lot of philosophy, especially mid and far eastern philosophy of the mind, the self, the five aggregates of Buddhism, etc. Psychology points at it, and modern psychology avoids it because it’s practically difficult for outcome oriented treatments.
>Generally though I challenge you to define “creative” in a way that is precise enough to measure and isn’t self referential or refer to concepts ill defined.
While I have no hope for a rigorous definition (I don't think it's possible), there are two very distinct kinds of creativity:
1. Result is sufficiently novel for the system itself, i.e. it never seen it previously. This kind is too trivial to even talk about.
2. Result is novel for the side observer. This kind of creativity is meaningless because it depends on at least one unknown (side observer).
the existence of free will is far from settled
https://www.smbc-comics.com/comic/agency