David Castillo

@smjain @rohanpaul_ai ...way more valid here, since it basically says "look, with everything we know we cannot discard consciousness", which I find very valuable in a scenario where everyone quickly jumped to say that there was no consciousness whatsoever. More "I don't know" is what we need.

@smjain @rohanpaul_ai I tried to read Penrose's hypotheses once and they all seemed like producing odd reasons to why human consciousness is so unique. To me it reminded me of the arguments people would make back when the mechanisms of life and reproduction were unknown. I think Hinton's message is...

"I don’t think there’s any reason why a machine shouldn’t have consciousness. If you swapped out one neuron with an artificial neuron that acts in all the same ways, would you lose consciousness?" ~ Geoffrey Hinton A fascinating discussion in this video. --- From the 'The Diary Of A CEO' YT channel

@elonmusk Ah, of course! way easier than convincing rich kids to stop going to the gym in their private jets. How didn't I think about this earlier?

A large solar-powered AI satellite constellation would be able to prevent global warming by making tiny adjustments in how much solar energy reached Earth

@VW_es Resumen: que me derivan a @vwcanarias Los mismos que me dejaron tirado en primera instancia...

@DCasBol Te hemos respondido por mensaje privado. Un saludo.

Acabo de aprender por las malas que si compras un eléctrico de @VW_es y tiene un fallo de fábrica en el motor/reductora estás VENDIDO. Ni 4 años y 50.000 kms que tenía el coche...

Ya disponible en @rtveplay el primer programa de UN NUEVO SENDERO HASTA EL MANANTIAL rtve.es/play/videos/un…

@jeremyphoward @clattner_llvm That question describes quite well how I felt while working with LLMs. And right now it's hard to find (funded) research topics that aren't buttered or splashed with LLM essence.

I spoke with @clattner_llvm about AI and software craftsmanship. "The question is, when things settle out, where do you as a programmer stand? Have you lost years of your own development because you’ve been spending it the wrong way?” A must read: fast.ai/posts/2025-10-…

@VW_es Enviado. Si de verdad es un caso aislado como me dicen en el taller oficial, no entiendo tanta reticencia a hacerse cargo. Como cliente me siento un cero a la izquierda...

@DCasBol Hola, David. Soy Alex, de Volkswagen. Sentiemos leer lo que nos comentas. Si lo deseas, estamos a tu disposición por mensaje privado, para que puedas ampliarnos los detalles de tu caso. Un saludo.

la reválida del #apagon #horaDelPlaneta es.wikipedia.org/wiki/Hora_del_…

#apagon en península, por fin llega la jornada de 6 horas!

@FonsiLoaiza Habrá que ponerle una velita a Luigi Mangione🙏🕯️

Hoy, Torbe ha vuelto a ser puesto en libertad sin medidas cautelares. Goza de impunidad en España. En Mediaset lo acaban de entrevistar en el programa de Juan Manuel Medina (padre de la nazi de Isabel Peralta) y Álvaro Rojo (hijo de Alfonso Rojo y Ana Rosa Quintana).

@sextaNoticias Lo único que necesita un país para sostener las pensiones futuras es no venirse a pique y garantizar recursos y servicios para las generaciones futuras. La cuestión monetaria nunca será un problema real, igual que no lo es para asuntos militares.

España necesita un millón de migrantes anuales para sostener las pensiones en las próximas décadas atres.red/q7r1a2

Se ha filtrado el borrador de ley de universidades que está preparando la CAM. Si se aprueba, será la muerte de la universidad pública madrileña. Es escandaloso. No exagero. Se vienen tiempos oscuros como no espabilemos. Explico algunos puntos críticos. Va hilo. 👇

@ylecun And actually I would expect this to work even better than normalization, given that it imposes a fixed range to most inputs in the network. And we all know what that means in terms of generalization!

New paper: turns out you can train deep nets without normalization layers by replacing them with a parameterized tanh()

@duncancampbell I don't know... but culture has a strong innertia. Unless enforced by governments or some major institution, I don't see that changing in the near future😅

@DCasBol Why haven’t we ever done that before in the US?

Somehow most people still have the flawed perception that changing the clock changes the sun dynamics🥲 the concept of having a single time zone and adjusting schedules locally simply escapes reason😓

@nickfloats ??? youtu.be/kY4lvLw41MI?si…

"Midjourney now has a a warehouse with hardware labs and 3D printers and CNC machines where we're building multiple things simultaneously. We're building both small hardware things that you can hold, and huge hardware things that you go inside of. We're preparing to bring in over a thousand gallons of, fluid soon for one of the projects. So we've got thousands of gallons of liquids coming in, and I won't say what kind of liquids, but it just sounds cool. Some of our stuff is starting to look a bit like CERN, like massive cables and power things and it's pretty fucking cool. And that's not even without the third major project under construction yet. But the third one's almost ready to start going. We have two that are basically physically producing things. Third one's gearing up and then there's a fourth one that we might fund. So I'm pretty excited about those things." -MJ office hours

@duncancampbell This is a very common misunderstanding, and I find it frustrating that most people don't get it: it's all a social agreement! e.g. in China they have no time zones, they simply let locals set their preferred schedules.

Very hard to understand why anyone would want this. Who cares about an hour of extra sun in the morning. But eliminating days where the sun sets before 5 sounds amazing. Don’t you want your kids to play outside after school? Sports, bike rides, family walks.

An Efficient Alternative to Self-Attention This paper presents FFTNet, a framework that replaces costly self-attention with an adaptive spectral filtering technique based on the Fast Fourier Transform (FFT). Key components: • Global token mixing via FFT – Instead of pairwise token attention, FFTNet uses frequency-domain transforms, cutting complexity from O(n²) to O(n log n) while preserving global context. • Adaptive spectral filtering – A learnable filter dynamically reweights Fourier coefficients, letting the model emphasize important frequency bands similarly to attention weights. • Complex-domain nonlinearity – A modReLU activation on the real and imaginary parts enriches representation, capturing higher-order interactions beyond linear transforms. Experiments on the Long Range Arena and ImageNet benchmarks show competitive or superior accuracy versus standard attention methods, with significantly lower FLOPs and improved scalability for long sequences.

but it's really good. It passed the anti-woke and capitalist tests 👌

I'm genuinely surprised by grok's performance. It doesn't show any of Elon's biases (which is a great relief). Actually, it does show little bias in many aspects! My only surprise so far has been seeing it sprinkle some English words in the Spanish responses 😅