ueaj

academics are unprepared for the coming world where much scientific progress is majorly a function of inference compute. whether OpenAI points the Eye of Stargate at your particular field will decide its acceleration. talent will leach away into the labs. it's already begun

NOW - Pope XIV says the church and Anthropic, will work together to "find the way for humanity, in this time of artificial intelligence."

update

UPDATE: According to sources who have seen the final text, Pope Leo XIV’s AI encyclical, set for release tomorrow, will argue that humanity faces a defining choice regarding AI. He uses the Tower of Babel — the ancient story of a people who tried to engineer their way to godhood and collapsed under the weight of their own power — as a warning about where unchecked technical mastery leads. Instead, the U.S.-born pontiff will urge the world to build something humbler in its place: a civilization where God and humanity can dwell together.

Google Cloud GPU% Any% Speedrun — World Record (18:54) [NO ORG POLICY GLITCH] — Quota Wall Skip Discovered!!

Extremely obvious that the bulk of replies to any post about the monumentally stupid green card change are bots or foreign accounts meant to make an unpopular policy that would strategically weaken the US appear more popular so as to increase the chance of weakening the US.

If you want to work on pretraining-for-AGI, join OpenAI, Google, Meta or the Anthropic/XAI/Cursor supergroup. The bitter truth of the widening compute gap is that all the problems which are actually on the critical path to AGI now demand that level of compute.

SITUATION DETECTED: Google DeepMind’s AI agent autonomously solved 9 of 353 open Erdos problems in mathematics, at a cost of a few hundred dollars per problem.

no I think it goes up and down forever. The only constant of our universe is emergent complexity. I think we'll discover (through simulation) that the quantum "noise" has structure, just like the minutiae of quantum chemistry appear as noise or exceptions to classical chemical laws.

simulating neurons in isolation falls short. we can't simulate brains yet for a reason. mitochondria as passive energy factory is exactly the kind of prevalent biological reductivism i'm challenging in this thread (and in most on neuro-vs-ai). for instance, mitochondria also happen to: create local energy gradients, regulate calcium, release signals that influence neuronal firing, and control cell survival and gene expression. those dynamics couple directly with firing probability, plasticity rules, and dendritic plauteau thresholds and abstracting them away in ur optimistic 5-line program would erase a wide web of control loops extending from what u and many engineers would assume a negligible passive power plant. this applies all the same to the assumption about neuronal behavior being sufficiently approximated by a ~dozen hidden states. the brain isn't a network of abstract point neurons reducible to simple spike-producing machines with a handful of states. their dendrites perform local, nonlinear computations, ion channels have rich history-dependent dynamics, and glia+neuromodulators constantly define how signals flow. the "extra stuff" u'd cut out isn't just decoration. the infra defines the computation. reproducing snapshots of neuronal activity would fall short in capturing counterfactuals for how those circuits would behave under largely different contexts. brains are noisy, analog, non-equilibrium systems. their noise, heat, and metabolic fragility aren't incidental infrastructure. to guarantee proper equivalence not just on today's inputs but across the organism's full operating regime, any emulator would have to also reproduce those nonequilibrium constraints, the stochastic fluctuations, the heterogeneity of cell states, and the multiscale feedback loops. once u include enough of that to ensure the counterfactual responses line up, the supposed dozen-orders-of-magnitude simplification collapses and u've essentially rebuilt a ~bio mortal computer under a different guise.