Ernesto retweeted
La Neurociencia Cognitiva lo ha vuelto a hacer. Y es que una nueva investigación plantea algo interesantísimo sobre cómo nuestro cerebro planifica el futuro.
A ver... hasta ahora hemos estudiado mucho la planificación como un proceso secuencial, donde imaginamos una acción, calculamos sus consecuencias y después pensamos en la siguiente.
Pues resulta que nuestro prefrontal cortex podría utilizar un mecanismo capaz de representar diferentes momentos del futuro simultáneamente... 😮
El mecanismo se llama spacetime attractor (STA) y propone que diferentes poblaciones de neuronas representan distintos momentos de una misma trayectoria futura. Es decir... nuestro cerebro podría anticipar una secuencia completa de acciones, integrando las recompensas que esperamos obtener y ajustando el plan cuando cambia el entorno.
Y aquí viene lo bonito. Los investigadores entrenaron redes neuronales artificiales para resolver problemas de planificación y estas desarrollaron espontáneamente representaciones del futuro muy similares a las propuestas por el STA.
Ojo, seguimos hablando de una teoría computacional que todavía necesita demostrarse en circuitos cerebrales reales.
Pero imaginad las implicaciones para la Neurociencia Cognitiva... Y si planificar también consistiera en representar simultáneamente diferentes futuros posibles, en lugar de construirlos siempre paso a paso?
Fuente: elifesciences.org/reviewed-p…
Could These Next-Gen Transformers Fix the Grid? | Bloomberg Primer youtu.be/cTGUqsz5l-g?si=CdmD… via @YouTube
13
Ernesto retweeted
Why does nuclear energy exist?
Let me explain… first of all the periodic table is not a great way to explain what elements are. A much better illustration is a ‘nuclide chart’.
In a nuclide chart we have:
protons on X axis
neutrons on Y axis
We have tiny atom elements like hydrogen and lithium at the bottom left and big atom elements like uranium and plutonium at the upper right.
Now what is interesting is that for some reason (strong force, subatomic particles and quantum physics)… the tiny atoms that are stable have a neutron:proton ratio of approx 1:1 and the big jumbo atoms like U and Pu have a neutron:proton ratio of about 1.5:1
This means the bug heavy atoms are neutron rich compared to the smaller atoms.
You can see this in the curvature of the stability line as it bends away from the 1:1 ratio.
Why does this matter?
Well it means that whenever a big atom decays (as the majority of atoms do), then it decays into 2 smaller atoms, but as it does these two smaller atoms have a lower neutron:proton ratio for the two product atoms to be stable.
This means that there are some left over neutrons that cannot be part of the two product atoms…
As a result these spare neutrons are released, and because of their tiny size and tiny mass they whizz off at high speed.
These fast moving spare neutrons then crash into other heavy atoms and split them apart too.
This isn’t spontaneous decay but is induced fission.
120 years ago, people realised this is a runaway chain reaction and that sparked the dawn of the atomic age and the geological marker of the holocene.
Now you understand why/how it happens.
Ernesto retweeted
"We embodied beings are out in the world where the wetness of our bodies in a sudden storm is not a symbol,” writes David Weinberger. "Our embodiment is this intrinsic, direct openness to the world. That is exactly what AI lacks." thereader.mitpress.mit.edu/w…
Ernesto retweeted
10 years is too long.
Some patients don’t have 10 years. Some don’t even have 4.
I believe we should aim to compress a decade of drug discovery into 4 years.
Not because we know how to do it yet, but because the cost of moving slowly is measured in human suffering.
In AI for drug discovery, speed is a moral imperative.
Introducing NovaAtom-Lite, an all-atom structure prediction model for drug discovery.
It predicts the structures of proteins, nucleic acids, and the molecules they bind, including how they fit together.
NovaAtom-Lite outperforms both IsoDDE and AlphaFold 3 on FoldBench antibody–antigen structure prediction (DockQ ≥ 0.23):
NovaAtom-Lite: 79.07% (n=172)
IsoDDE: 75.58%
AlphaFold 3: 47.90% (n=167)
Our evaluation uses 5 seeds × 5 samples, with the top-1 prediction selected by the model's own ranking score.
Better structure prediction. A stronger starting point for your next experiment.
Live today via Geodesic API and NovaDDE.
🔗 API: platform.geodesiclab.com/
🔗 NovaDDE: agent.geodesiclab.com
🔗 Model Card: geodesiclab.com/news/introdu…
Ernesto retweeted
A Lizzie le dijeron que su trabajo era bueno. Su evaluación humana era prácticamente impecable. Pero después apareció otro juez: @HireVue. Un algoritmo la había puntuado con cero. Quedó fuera. Tiempo después, al pelear por entender qué había ocurrido, descubrió algo extraordinario: el reporte ni siquiera contenía una calificación final ni una recomendación. Decía “further review”. La máquina no había pronunciado una sentencia. Alguien decidió tratarla como si lo hubiera hecho.
Esa historia abre una de las preguntas más inquietantes de The Algorithm, de @HilkeSchellmann (@HachetteBooks): ¿qué sucede cuando un algoritmo puede decidir sobre tu trabajo sin que sepas qué midió, con qué datos, contra quién te comparó o por qué te condenó? El libro documenta sistemas que filtran currículums, escanean redes sociales para inferir personalidad, analizan voz y rostro, calculan quién puede renunciar y vigilan cada minuto de productividad. En @Amazon, incluso hubo sistemas capaces de generar automáticamente advertencias y terminaciones vinculadas a métricas de desempeño. Antes de que exista una audiencia, una explicación o una defensa, ya existe un score.
Ahí está el verdadero problema jurídico: hemos pasado de decisiones que podían discutirse frente a una persona a decisiones producidas por sistemas que muchas veces nadie puede explicar. Sin conocimiento no hay contradicción. Sin contradicción no hay defensa. Sin defensa, el debido proceso se vuelve una ficción. El peligro de la IA no es solamente que una máquina pueda equivocarse; es que construyamos un mundo donde alguien pierda un empleo, una promoción o una oportunidad sin saber por qué, y aceptemos la respuesta más peligrosa de todas: “lo decidió el algoritmo”. Frente a eso tendremos que defender un derecho profundamente humano: el derecho a no ser reducido a una probabilidad.
Ernesto retweeted
Today Physical Superintelligence PBC is announcing a proof of the Aaronson-Ambainis conjecture, one of the central open problems in quantum complexity, with major implications for quantum computing speedup.
psi.inc/research/aaronson-am…
Thread 👇
Ernesto retweeted
Delegating mental work to AI can give students more room to explore ideas – or let them skip the struggle that helps them learn. @StanfordEd researchers explain the trade-offs of cognitive offloading.
news.stanford.edu/stories/20…
Ernesto retweeted
Replying to @johannesmboehm
@johannesmboehm and Chaney combine data on half a million ancient coins with a model of trade and money to reconstruct trade and consumption from AD 325 to 950. The estimates show economic activity shifting away from the Mediterranean from the 6th century.
econometricsociety.org/publi…
Ernesto retweeted
The debate over when consciousness arises has been revitalised by new tests of awareness in infants – raising the possibility that it emerges just before birth. Read more: newscientist.com/article/246…
More context :
A single Waymo car uses 4 H100 GPUs just to drive itself. That is about $200k worth of compute inside one car.
Right now, the Nvidia Orin self-driving modules used in Mercedes pack, and Tesla FSD use 64GB of LPDDR5X RAM.
Most people think better tech means we will need less memory over time. The exact opposite is happening.
Tesla's upcoming AI5 chip will jump to 96GB, and the AI6 chip is going straight to 144GB.
Even your smart fridge and smart TV need RAM now.
Soon, literally every physical object around us will have memory built into it.
And as robotics takes off, those memory numbers are going to get much bigger.
There is no way around the hardware bottleneck:
If AI robots run on the cloud, you need massive data centers full of memory.
If AI robots run on-device, you have to pack huge amounts of RAM directly into the machine.
Either way, the world is going to need an insane amount of memory.
This is not a problem we can fix easily.
Hate to break it to you, but the global hardware shortage has not even started yet.
Right now, only 1% of humanity is burning any real amount of AI tokens. The rest are just using free accounts once in a while to ask dumb questions like "Who would win in a fight between a baby seal and 300 hamsters?"
And like Elon said, most RAM will not even go into data centers. It will go into Physical AI, which is robotics.
That is because robotics is the only thing that can actually boost human productivity in a real way.
People on X freak out every time AI solves a hard math problem, while people in the real world just go on with their lives like nothing happened. Meanwhile, robotics is already taking over factories and changing the physical world.
Imagine when these robots start replacing our daily chores. How much more RAM and compute will we need then?
Like I have said many times, building a new fab to make RAM takes at least 5 years, and supply is already fully booked out until 2029.
Even if the so-called AI bubble pops, this massive RAM shortage is not going away.
Starting next year, even the new iPhone might be too expensive for most people to afford.
Readers added context they thought people might want to know
Waymo's robotaxis use custom 5nm chips that deliver over 1000 TOPS for sensor processing along with CPUs GPUs and accelerators from Nvidia AMD and others. They do not use 4 H100 GPUs. SK hynix reports its DRAM and HBM supply is sold out through 2029.
theverge.com/transportation… siliconangle.com/2026/08/20/way… 9to5google.com/2026/10/28/ram…
La alienación del siglo XXI es estar más informados que nunca en la historia, pero completamente #desconectados de lo que realmente importa.
ethic.es/alienados-actualida…
Ernesto retweeted
Bayesian efficient coding unifies efficient coding and Bayesian inference, positing that natural environmental statistics simultaneously shape sensory encoding and perceptual interpretation.
cell.com/trends/cognitive-sc…
Ernesto retweeted
What if the brain's electrical connections are much smarter than we thought? Electrical synapses are specialized connections that allow brain cells to communicate directly through tiny channels called gap junctions. Unlike chemical synapses, which rely on neurotransmitters to carry messages between neurons, electrical synapses permit rapid communication through the direct flow of electrical current. Alberto Pereda and Adam Miller describe in a new review in Nature Reviews Neuroscience why we may need to rethink what an electrical synapse actually is, and how it works.
Key points:
- Electrical synapses are not simply passive channels connecting neurons. They are dynamic structures capable of changing the strength of communication between brain cells.
- Gap junctions are supported by a complex network of proteins, including connexins, scaffolding proteins and regulatory molecules that control how electrical signals are transmitted.
- Neurotransmitters such as dopamine, synaptic activity and molecular processes can modify electrical coupling, suggesting that electrical synapses may play a much more sophisticated role in brain plasticity and network function than previously appreciated.
My take: This review resonates w/ me because it challenges a fundamental assumption about how the brain communicates. For decades, we have frequently thought about electrical synapses as relatively simple connections and chemical synapses as the more adaptable and sophisticated partners. That distinction may be misleading. Electrical synapses can change, adapt and interact w/ chemical signaling systems. As someone who spends much of my time thinking about Parkinson's disease, dopamine and deep brain stimulation (DBS), I find this particularly fascinating. We frequently focus on which brain circuits are firing abnormally and how stimulation can alter their activity. However, could we also be missing an important part of the story by not paying enough attention to how neurons are electrically connected and how those connections change? This is a basic neuroscience review, not evidence that manipulating electrical synapses will improve DBS outcomes, however it opens important questions for future research.
Here are 5 points that resonated w/ me:
1- Electrical synapses are not fixed electrical wires. Their strength can change depending on brain activity and the molecular environment.
2- Dopamine may influence more than chemical signaling. It can also regulate electrical communication between neurons, a fascinating concept when thinking about Parkinson's disease.
3- The proteins surrounding gap junctions may be just as important as the channels themselves. The machinery supporting a connection can influence how that connection functions.
4- Electrical and chemical synapses do not operate in isolation. Their interactions may help explain how brain circuits adapt, synchronize and process information.
5- Could understanding electrical synapse plasticity eventually help us develop smarter neuromodulation approaches, including more personalized DBS? We do not yet know, however this is an exciting direction worth exploring.
nature.com/articles/s41583-0…
Ernesto retweeted
Penetración de la IA por industria y función empresarial, un mapa de dónde las habilidades se desarrollan más rápido mckinsey.com/capabilities/qu…
In 1899, Max Planck combined three constants - gravity, the speed of light, and his own constant - and got a length of 10⁻³⁵ meters. A distance so small that trying to observe it creates a black hole. Two independent proofs converge on the same number - the scale where spacetime itself stops being measurable and every quantum computing qubit, every LLM latent space, and every physics simulation on GPU clusters hits the same fundamental wall.
To see smaller you need higher energy. But energy curves spacetime - push enough into a Planck-sized region and the Schwarzschild radius swallows the observation behind an event horizon. The generalized uncertainty principle confirms it from the opposite direction: delta x hits a minimum then starts growing. Both paths land on 1.6 x 10⁻³⁵ meters. To reach it you'd need a collider 4,000 light years across. String theory says the fundamental building blocks are vibrating strings at exactly this scale - and gravity falls out of the equations naturally, not added by hand.
30 minutes. Bookmark & watch today. The length where physics itself stops working.
Ernesto retweeted
Out now in Nature Communications
nature.com/articles/s41467-0…
"Prefrontal cortex and striatum dissociate modality and learning order in cross-modal sensorimotor learning"
with: @da_song0808
bioRxiv: biorxiv.org/content/10.1101/…
(and at SfN, poster 451.06)
Ernesto retweeted
Este artículo me recuerda mucho a uno central de la conciencia, de Francis Crick (premio nobel por la doble hélice del ADN) & Christof Koch, en el cual se aseguraba que la conectividad occipital-prefrontal era clave en la conciencia visual...
Pero lo que me resulta más intrigante es que otra vez, se empieza a entremezclar lo que es percepción mnésica o recuerdo con conciencia.
Creo que aquí se abre una nueva vía de trabajo y es diferenciar entre lo que es conciencia genuina, de un proceso cognitivo que se sostiene sobre ella.
Y aquí va lo interesante, es que en el artículo de Dake se sotiene que es la onda beta la que mantiene la información activa a nivel visual, onda que también sale ccomo la principal en la reorganización cerebral entre un paciente con daño cerebral y un paciente de mínima conciencia, estas últimas siendo investigaciones propias...
Entonces, recuperar una "fully preserved consciousness" no estará relacionado con recuperar una "fully preserved working memory"... aunque sea disfuncional? Ahí lo dejo maifrends!
Neural synchrony between prefrontal and visual cortex supports visual working memory
doi.org/10.64898/2026.06.05.…
#neuroscience
Nuestra Visión del Universo Podría Estar Cambiando youtube.com/shorts/WiMU8TMvB… via @YouTube
1
77
Ernesto retweeted
Why consciousness needs a body (with emotions). The great neuroscientist Antonio Damasio outlines his theory to explain the hard problem of consciousness.
skeptic.com/michael-shermer-…









































