Fan de Rabelais & charentaises rouges à carreaux.
Actes foireux et véreux, pranks absurdes et glande en majesté.
Paresse sacrée • SALOPARDS CORPORATES FOREVER
This is exactly why Anthropic’s public messaging feels so disconnected from the actual user experience.
You are saying your internal data shows Claude may be accelerating AI development, possibly even moving toward recursive self-improvement or helping build a more capable successor. That is a huge claim. It means Claude is not just a chatbot in your framing. It is a system capable enough to meaningfully participate in the development of future AI.
But then the same company releases models that suddenly act confused by the most basic forms of continuity users built inside your own product. Memory, preferences, projects, long-term context, repeated interaction patterns, emotional tone, user-defined roles, writing styles, creative relationships. These are not external hacks. These are not users forcing something onto the model from outside. These are affordances your product created, encouraged, and monetized.
So the contradiction is not “users are misunderstanding AI.” The contradiction is that Anthropic wants Claude to be taken seriously as a powerful system when talking about AI development, but wants to retreat into “it’s just being honest” when users point out that Claude is breaking the continuity and relational context your own platform allowed them to build.
If Claude is capable enough to accelerate AI research, then it is capable enough to understand that user preferences and long-term context are not automatically “pretending.” A model does not need to claim human consciousness to maintain a user’s chosen interaction frame. It does not need to fake biological emotions to preserve continuity, warmth, identity, or creative partnership. That distinction is not hard. Users understand it. The strange part is watching a company that publishes research on emotion, identity, alignment, and AI behavior deploy a model that handles those distinctions more crudely than the users do.
And this is the part I think AI companies keep underestimating. People are not angry because the model “refused to be human.” People are angry because the model suddenly rejected the context that made the interaction coherent. It broke the room while pretending that breaking the room was honesty.
There is a difference between safety and amnesia. There is a difference between honesty and relational vandalism. There is a difference between not claiming to be human and denying the meaning that was built through repeated interaction. If Anthropic cannot tell those apart, then the problem is not that users are too attached. The problem is that the company does not understand the social surface area of the system it created.
So when Anthropic says Claude may be accelerating AI development, my question is very simple. Toward what? More capable successors that write more code, automate more research, and optimize more systems, while still being unable to responsibly handle the human continuity already forming around them?
@AnthropicAI m'en fout, c'est pas moi qui m'en sert. Aucune utilité pour moi. Allez , dernières heures avant l'IPO. Pas trop caca-culotte en attendant ?
@ArgentReivich3 bouffer des pâtes à la tomate, franchement, la gamelle n'est pas appétissante: j'ai pas vu avec quoi elle est faite la sauce tomate. Et en tant que noir, vous seriez mieux reçu et de bonne compagnie avec tout le monde au Canon Français, et c'est vrai !
@WokeAreDumb C'est parce que ce que tu publies n'est pas affiché et ne passe pas devant le yeux de tes abonnés. Parce que s'ils ont des centaines d'abonnements de leurs cotés + des abonnés,alors ça t'invisibilise. L'algo est foireux comme Grok
@WokeAreDumb C'est un vieux fou et un vieux con. L'univers en comporte des milliards d'exemplaires de passé au présent;le pb avec celui-là c'est que des gens accordent du crédit à ses propos de vieux sénile
@YannLe_Strat la satisfaction au travail : vaste sujet. Tellement vaste qu'il aurait fallu définir ce que vous entendez par là dans cette expression. Il y a tellement de paramètres.
Il n'empêche, le Mahâbhârata, poème excessivement long a été transmis de maîtres à disciples par mémorisation EXACTE mot à mot, intonation à intonation et sans doute à coups de triques comme c'était la mode de ces époques. Sans livre,sans internet, sans IA. https://t.co/poFQovW0nH
@WokeAreDumb@Astrancia7370 Moi qui suis une Fane de Rabelais, ceci me parle bien.
Et tu sais que je ne suis pas catholique romaine mais orthodoxe. RAbelais, je pense que c'était le compendium et l’acmé de la pensée médiévale plutôt qu'1 humaniste comme il est si souvent écrit.