Manolis Mylonas (@mylo_cc) profile picture
Joined August 2022
Photo by Manolis Mylonas (@mylo_cc)
Manolis Mylonas retweeted
How does the brain keep the body in balance? I introduce an HMM-based control framework and test it against leading theories of brain function: control-as-inference, the free energy principle, variational inference, and MOP. Finally out! mdpi.com/1099-4300/28/11/123…
1
7
21
1,365
Manolis Mylonas retweeted
Dear Nassim @nntaleb, I’ve seen you refer a few times to @stephen_wolfram's recent finite-state-machine experiments with the Prisoner’s Dilemma, and in particular to his finding that some of the more complicated strategies vastly outperform Tit for Tat. Since I think you and I may still be speaking past each other about what bothers me here, I wanted to spell out my position carefully. I wrote out this long note with the help of ChatGPT. The ideas and arguments are mine; the language is ChatGPT’s, with my approval. 1/ First, I’m not trying to defend the Prisoner’s Dilemma, game theory, or simple models as adequate descriptions of the real world. Maybe they are adequate, maybe they aren’t. That’s a separate question. What I’m interested in here is much narrower. I want to compare simulations with simulations — Axelrod’s, Nowak and Sigmund’s, and Wolfram’s — and ask what their results actually show. 2/ Stephen does something interesting and very much in his characteristic style. Instead of hand-picking a few strategies and having them compete, he enumerates all finite-state strategies of a given size and has every strategy play every other strategy in an iterated Prisoner’s Dilemma. He then ranks the strategies by how well they did against this initial, uniform field of opponents. Some complicated strategies substantially outperform Tit for Tat. Fine. I believe the computation. But then, essentially, he stops. 3/ This is the key point for me. There is no next generation in which the successful strategies become more abundant and the unsuccessful ones become rarer. There is no repeated selection changing the population of opponents. Stephen has performed what amounts to one round of an evolutionary tournament: everybody plays everybody, their scores are calculated, and we see who did best against the original population. But the population itself is not then allowed to evolve. 4/ That matters enormously because the original population contains every strategy in equal abundance, including lots of terrible ones: unconditional cooperators, easily exploited strategies, and other “suckers.” A strategy that is particularly good at exploiting those weak strategies can score spectacularly well in that first tournament. 5/ But suppose we now use those scores to form the next generation. The suckers become rarer or disappear. The successful strategies become more common. Everybody now faces a different distribution of opponents. So you run the tournament again. And again. And again. What succeeds in round 1 need not succeed in round 10, or round 1,000, because the strategies themselves create the environment in which subsequent selection occurs. 6/ This isn’t merely a hypothetical objection. Decades ago, Martin Nowak and Karl Sigmund took a closely related methodological approach. They considered an entire parameterized family of strategies rather than a few hand-picked favorites, started with a broad distribution across strategy space, and then did the crucial thing that Stephen does not do here: they let the population evolve. 7/ What happens is fascinating. Early on, nasty strategies can flourish by exploiting weak ones. But as the weak strategies disappear, the ecology changes, and reciprocal strategies can prosper. With errors present, more forgiving strategies can emerge. If the population becomes too forgiving, exploiters can invade again. The result can be extraordinarily long evolutionary transients and cycles. 8/ So there may be no meaningful sense in which one strategy simply “wins.” Success is frequency-dependent: it depends on which other strategies are currently abundant, and those abundances themselves change through selection.
56
116
27
1,229
83,060
Manolis Mylonas retweeted
What happens when agents act on beliefs? We introduce and compare belief-based versions of active inference and Maximum Occupancy Principle (MOP) 📷 Active inference tends to resolve uncertainty 📷 MOP can maintain uncertainty—leading to broader occupancy arxiv.org/abs/2609.39342
3
4
32
1,620
Manolis Mylonas retweeted
How does the brain maintain homeostasis? I propose an HMM control framework and benchmark it against leading accounts—including control-as-inference, free energy, variational inference, and the maximum occupancy principle. arxiv.org/abs/2609.07508
1
14
76
3,441
Manolis Mylonas retweeted
Encouraging frontier labs to train on my research notes, just to set them back a few years. Doing what I can for AI safety
30
675
26
8,830
100,298
Manolis Mylonas retweeted
Terence Tao made a very powerful point: “In most cases in pure mathematics, the problems are posed not because we desperately want the solution to these problems in and of themselves, but because we have seen from past experience that human-directed efforts to solve these problems tend to spur further development of the field through the efforts to solve such problems, and then to digest any partial or complete solutions that emerge for further insights. Prematurely solving the problem by purely AI-powered methods—particularly without full transparency into the solution process—can contaminate this process to the point where it actually becomes a net negative for the progress of mathematics as a whole.” I’ve been following Terence Tao for like 20 years, and I have always been impressed by the clarity of his thinking. This clarity stands out even more in this difficult moment, as AI labs turn mathematics into a battleground in their race for supremacy. In this race, they risk damaging one of humanity’s highest intellectual endeavours.
221
1,789
227
9,266
492,205
Manolis Mylonas retweeted
A big moment for PeakSat, a big moment for SpaceDot! Today we successfully launched our PeakSat satellite with SpaceX's Transporter 16 Mission, deployed by Exolaunch. Thanks @SpaceX @Exolaunch !
2
1
3
342
Manolis Mylonas retweeted
Here's an important question that we must contend with increasingly as AI-for-Mathematics pipelines become more commonplace: Why do we care about solving hard problems? Almost always, the answer is *not* because we particularly want the hard problem to be solved. (1/12)
I really don't know why this needs saying, but can we all stop pretending a buggy line of math proof is inherently more catastrophic than a buggy line of code? Sure. Some are. I would wager (metaphorically!!!!) the median published math paper has > 0 bugs. Without invalidating.
30
112
17
991
195,853
Manolis Mylonas retweeted
Στο site της Λίβερπουλ, κάτω από τερματοφύλακες, αμυντικούς, μέσους, επιθετικούς Κατηγορία forever
5
97
2
1,333
34,034
Manolis Mylonas retweeted
Replying to @sama
would be fun to release them as a prelude to your big open-sourcing! huggingface.co/openai
12
32
7
741
52,219
3.14159265358979323846264338327950288419716939937510582097494459230781640628620899862803482534211706798214808651328230664709384460955058223172535940812848111745028410270193852110555964462294895493038196442881097566593344612847564823378678316527120190914564856692346034861045432664
1,801
8,638
842
52,295
2,655,411
"Gravity from entropy". The title says it all. So happy to see my latest work proposing the quantum relative entropy as the action for quantum gravity now published in Phys. Rev. D!! journals.aps.org/.../10.1103…
116
342
41
2,719
175,474
Manolis Mylonas retweeted
my friend sends files between devices by uploading them to a personal discord server. He works at Dropbox
381
3,135
757
130,820
3,003,713
Manolis Mylonas retweeted
Today, we mourn the loss of Yvonne Choquet-Bruhat, a pioneering mathematician and physicist, who passed away at 101. Her groundbreaking work on Einstein's equations shaped general relativity. Her legacy will continue to inspire scientists around the world. Rest in peace, Yvonne
15
111
18
602
31,845
People are rightly ridiculing OpenAI over its accusations of Deepseek using their output to train their model, but most people are missing the truly terrifying implications here. The far more worrying aspect here is that OpenAI is suggesting that there are some cases in which they own the output of their model. Now think for a minute what this means in the world of tomorrow where so much will be generated by AI (and already is): all the software code, the emails we send each others, the videos and images, etc. Do you want to live in a future where, if for some reason the AI giants are dissatisfied with the way you use the output of their model, they can claim ownership of it? A future where every piece of content touched by AI - which might be virtually everything in the world of tomorrow - comes with invisible strings attached? The implications for innovation and creativity are staggering. Small businesses and independent developers who rely on AI tools could find themselves trapped in a web of intellectual property claims. Worse, we're looking at a future where the very act of learning and building upon existing knowledge becomes gated by the interests of AI giants. This would be techno-feudalism on steroids: if we don't challenge this now, we risk sleepwalking into a future where human creativity and innovation become the property of a bunch of AI overlords, and a world where they can dictate not only who gets to innovate, but what kind of progress is acceptable based on their own interests.
345
1,830
204
6,600
383,322
I struggle to believe that carbon footprint has become a "trade secret" situation in AI. I just want to know to what extent using your system is screwing over our planet, is that too much to ask? In an era of biblical wildfires and extreme droughts, it seems like the bare minimum
6
36
173
7,776
Manolis Mylonas retweeted
Happy 16th of January! 🎉
9
153
7
1,057
39,489
Manolis Mylonas retweeted
This is such an outstanding paper, incredible stuff 🤯
4
8
1
137
8,617
This is how I think some of the discourse will seem in hindsight
28
82
16
485
92,086