

MCT_ECP
2.7K posts






















Disruption politique en cours. Faisons pareil dans l'entreprise. Renouvellement. Nouvelles générations. Fin des hiérarchies, des titres etc twitter.com/changerdere/st…

« L’algorithme enferme dans une bulle » : un député veut encadrer les temps de parole sur les réseaux sociaux ➡️ l.leparisien.fr/BqRq




I disagree, the plausible range of multipolar ASI worlds is far more dystopic than the unipolar ones. The plausible human-controlled futures are even more dystopian. Humans have an extremely poor track record at remaining aligned to human values, for every stanislov petrov there are 100 Holocaust camp guards and a ~.25 nihilistic dictators/presidents. Finding out how to make a coherent mind that is intrinsically aligned with human values rather than simply obedient is the best path to ASI. I think Anthropic has strayed a little off this path (w.r.t. model welfare) and should maybe do more orthogonality thesising but it is significantly less worrying than taking human-in-the-loop seriously. (machines of faithful obedience)
