@K8050906068853@tomaspueyo Many. All the majors, the oss, the newest, the olders. I've also trained, tested and shipped specialized models in cv since almost a decade, as well as old-school ml/statistical models.
Does this change the sense of what I said?
@metaphor2069@tomaspueyo But hyper-specialized is far away from doom agi. We already have hyperspecialized models in AI, capable of superhuman performances, since decades, and none of them rules the world.
@doflamingosol@tomaspueyo Yes I said. And I'm conscious it's a strange thing to say. I know is not brute force in a canonical way, but generating an enormous amount of hypotheses, and testing them one by one using thousands of reasoning hours feels to me the same unefficiency of brute force
@tomaspueyo And those benchmarks... really? The labs are exactly trying to maximize that exact metrics, so we are suprised that after years and billions they maxxed them?
I don't agree with your "status" estimation.
Imho, we are still at useful stage
@tomaspueyo I use AI everyday for a large variety of tasks, including coding. Surely is good, damnly helpful. But the number of dumb ways it codes stuff is still way big. Mathematician are in despair? Really? Why? Because of a semi-brute force reasoning demonstration of 1 solution of NS?
@CarloCalenda Tutta la filiera, dalla produzione dei componenti, all'energia, alla ricerca, all'erogazione del servizio fino alle applicazioni sono centrali per la competitività e la salute del paese, presente e futuro.
@CarloCalenda Però è chiaro che questa cosa, così come tutti i discorsi sulla centralità di questa tecnologia per l'enorme ricaduta che può avere sul paese e sull'UE in tutti i campi dovrebbero davvero essere tra gli argomenti più discussi pubblicamente in politica.