Sabitlenmiş Tweet
happy to announce i just got examiner feedback on my thesis. recommended for acceptance as is, no changes, no corrections :)
English
Michael Timothy Bennett
6.9K posts
Michael Timothy Bennett
@MiTiBennett
award winning ai researcher (2 so far) | just graduated cs phd | it is my laptop now | author how to build conscious machines | musician | @bennettsrazor
Brisbane, Australia Katılım Şubat 2020
4.7K Takip Edilen14.9K Takipçiler
Dropped a table on my foot. Broke middle toe (again - it used to get caught in the links of the kick pedal chain when I was a drummer). But I’ve been making progress running, so I kept at it. Turns out a broken middle toe is about the least painful thing to break?
English
Michael Timothy Bennett retweetledi
@joyal_kenus @MiTiBennett I have the hardcopy too! There's just no replacing the feeling of a real book in your hands, especially for a read this thought-provoking.
English
mmmm, quadrillion dollar research
Elizabeth Wood 🧬🖥️🥼@lizbwood
Our paper on variational synthesis is out now in Nature Biotechnology. Manufacturing-aware generative models — AI architectures that know how to physically build their own designs — enabling synthesis of DNA encoding ~10^16 AI-designed proteins at a cost that would be roughly a quadrillion dollars using conventional methods.
English
@borisyeltzin @sebkrier you overestimate how hard I was working before
English
@MiTiBennett @sebkrier "i've told you a million times not to exaggerate" - Rick, The Young Ones
English
Feels a bit like asking Twitter what drives most engagement. If your testing medium is short form videos, you're measuring what performs well in that format, which structurally rewards emotionally charged, high-contrast messaging.
"Within 5 years, Al is projected to eliminate 75% of our jobs" is false, but yes I'm not surprised this maximises engagement. Doesn't mean it's worth degrading the broader information environment further though.
David Shor@davidshor
When we tested videos on these themes, AI-specific populism - messages that address the future impact of AI in bold, populist terms - performed at the top compared with every other topic we've tested in moving voters toward Democrats.
English
Michael Timothy Bennett retweetledi
CFP for ALIFE2026
What is life, what is life-like? This special session explores artificial life as natural philosophy: what computational models offer for philosophical inquiry, and what philosophy looks like when you have to build it.
3–8 pages, due March 30!
English
It's funny; just realized I don't actually have a printed copy.
Joyal kenus@joyal_kenus
my read for the week 💯 Ty @MiTiBennett
English
Michael Timothy Bennett retweetledi
@MiTiBennett @elijaperrier My Claude can’t keep up reading all these man cmon
It’s ya boi
English
Thanks! I can definitely see a "keep your options open" connection, but the formal objects differ in an important way. Empowerment = channel capacity (bits of perceivable causal influence). Weakness = extension cardinality (how little a policy commits to). You can have high empowerment with a non-weak policy, and max weakness with zero empowerment. The main advantage of weakness is proofs that it is necessary and sufficient for optimal generalisation under a maximally uninformative prior (or for any prior under an additional set inclusion constraint). Started as a way to avoid the interpreter dependent performance issues associated with complexity based induction systems in interactive settings.
I think there are some interesting connections between our work more broadly. Happy to chat over email if you'd like?
English
This looks really interesting!
After a quick read — I think your concept of maximizing policy "weakness" strongly resembles a discrete, unweighted analogue of "empowerment" (a measure of agent's potential control over its environment: ieeexplore.ieee.org/document/15546…)
Have you considered bridging Stack Theory with the empowerment literature? It seems like an natural extension
English
@AgustinMIbanez @atlanticfellows @BrainlatUAI @Psicologia_UAI @ReDLat_Dementia @TCINeuroscience @CNCUdeSA @LACCD9 @tcddublin @UAI_CL @gerochile You might find this interesting. The chord / arpeggio thing is from my earlier workshop paper on machine consciousness (the somewhat dramatically named “A Mind Cannot Be Smeared Across Time”), but this develops it a lot more for consciousness in general
Michael Timothy Bennett@MiTiBennett
English
Music helps to understand the mind and the brain. Throughout the history of science, metaphors have shaped how we understand complex phenomena. The brain-as-computer metaphor has guided decades of theories and research. We propose music as a scientific metaphor for understanding the mind and brain via triplicate interfaces (listener, performer, composer) and a compound set of predictions. Multiple domains of music can be mapped onto different neural, cognitive and intersubjective processes such as network coordination, prediction, emotion and meaning. Neurocognition is not static but a dynamic, embodied, and time-sensitive system, much like a self-organized orchestra in which multiple processes interact simultaneously. Drawing on synergetics, predictive processing, and embodied cognition, we outline musical principles illuminating cognitive and action integration across time, offering new conceptual frameworks and testable predictions for future research. I enjoyed writing this piece with these stellar authors: @Kaiameye, @acolverson1, Christopher Bailey, @brucemillerucsf, @dafneduron90, Nicholas Johnson, Olga Castaner, @PierLuigiSacco, Eoin Cotter and Lucia Melloni. Science, like music, advances through new ways of listening to complex systems: doi.org/10.1016/j.neub…
English
@pfau you mught find this interesting
Michael Timothy Bennett@MiTiBennett
I am immensely proud and grateful to announce that, after three years of hard work, my paper on biological vs artificial intelligence has been accepted to the oldest and longest-serving scientific journal in existence: Philosophical Transactions of the Royal Society B.
English
We need a grand unification between physics and computer science to understand the relationship between energy and information. Always nice to see work that brings them together.
Hadi Vafaii@hadivafaii
The "decoupling of information and energy" is a major point of divergence between biological and artificial computers. Brains are efficient, modern AI isn't. And energy consumption is the biggest bottleneck in scaling AI (you can't hallucinate electrons into existence). To address this we need an "energy-aware theory of computation." And this new preprint is an attempt to address this. [1/11] 🧵
English
@chuswlove His books and my subsequent conversations with him have had quite an influence on my work! I even cite his thesis on harbour seals in an upcoming paper
English
In Peter Watt's Blindsight, a faked experiment shows how a digital mind experiences consciousness at lower and lower time resolutions, even at the scale of seconds. I wish Michael's work would have been in the Annex!
Michael Timothy Bennett@MiTiBennett
English
I don't see how. The question is why does life exist. To get to that, I ask what exists. I set up a formal model I point to two sorts of things that arise in this model; complex but weakly constrained, and simple. The why of it is just that everything that can exist will across envinrments, and anything but those two sorts is going to get deleted. The vulnerability isn't the "why does this exist" part; that's just some math within the model. The vulnerable part is whether people find the argument that the math represents what I say it represents compelling. i.e. that complex + weak is life, or that what I've formalised is properly analogous to the undefined functional information. The Functional Information people already argue much of the other stuff.
English
