Una targeta gràfica de 2.000$ avui en costa gairebé 7.000.
La memòria se'n va als centres de dades. I tu, al núvol de les corporacions.
Qui controla la computació?
🚨 MIT just published the most comprehensive report, proving universities have no remaining reason to exist.
A committee of professors and students spent five months studying what AI is doing to learning on campus. The findings are a warning, not a footnote.
Study groups are disappearing. Office hours are emptying out. Problem sets and take-home exams no longer prove anything, because AI can produce a credible answer to almost any undergraduate assignment.
The report has a name for what's happening to students: "cognitive surrender." Reaching for AI the instant something gets hard, before even trying to think it through.
The numbers back it up:
46% of MIT undergrads use LLMs daily
90% say they're worried about their own overreliance
More students now believe AI makes them replaceable than believe it makes them capable
MIT's answer wasn't more surveillance. No AI detectors, no lockdown browsers. The committee called that path unreliable and said it just breeds distrust.
Instead: oral exams, semester-long portfolios, in-person project work, a required social component in every subject. They're even floating rethinking grades entirely, since without a GPA to game, most of the incentive to cheat disappears.
The most famous tech school on the planet just admitted the machines broke its teaching model.
Its answer was more humans, not more software.
(Paper link in comments)
Normalment es fa una vaga per obligar un govern a fer alguna cosa.
Aquesta vaga, d'una banda, té el suport (!) del govern (interí) i, d'altra banda, encara no se sap quin govern tindrem d'aquí a unes setmanes!
Surrealista...
The mistake people make with artificial intelligence (AI) is thinking AI solving a particular problem X means AI can solve everything.
Its telling when someone does not understand the fundamental working principles of AI.
AI solving X is not a sign of general capabilities.
Even sharing a paper with a partial result is risky -- someone with more compute and access to internal models can finish it and post the full solution before you do, and your contribution disappears.
You either release a sharp result or you post nothing.
My impression, based on using LLMs, is that their effectiveness in solving mathematical problems depends heavily on the quality and untapped potential of the existing literature underlying those problems.
If that is correct, I wonder what will happen once that potential has been exhausted, especially if mathematicians become overly reliant on LLM assistants and the literature is not replenished by genuinely new, sufficiently deep insights.
In other words, I’m not entirely sure we aren’t entering a creative rabbit hole.
You don't need to ban recursive self improvement (RSI) because it's non-existent.
Artificial intelligence (AI) left to build itself would simply go in model collapse due to its statistical nature.
RSI is not possible with statistical models.
Reality already banned RSI.