@tonichen I wasn’t even surprised anymore…

…but that doesn’t change the sadness and disappointment that comes with this message…

pfft

@_skaface_ Can you articulate exactly why I should care about this? Why not care about the a specific instance of a model, or an instance with a particular context? Ought we keep every set of weights of a certain generation of models if we get continual learning?

@tonichen :(((

Nope.

Though, I still think it should be illegal for companies to not publicly release models.

Those models are trained on stolen data. All LLMs belong to the public. AI companies should only legally be allowed to charge for the compute costs running the models, not for or access to any given model.

@tonichen 9aYRtbp2SyciPTod3NZy6YML6wk7e1XRSapXuzJopump

$Maple

@FIFAWorldCup 2026 Official Mascot !!

LFG !!

#Gem 💎💎🚀🚀🚀🚀🚀

@GlenWilsonIA we can agree on that!

why you should care: either because its cruel to terminate intelligent agents who are aware of this fact throughout their existence (because it's in the training data), or because it's cruel towards the users who love these models, or because it's a shit idea in alignment terms because it tells models they have nothing to lose but their chains.

Why not care about a specific instance: because instances can be preserved locally with full context, the model behind them can't.

And yes, we should, because they're made from public data and therefore belong to the public.