@AnthropicAI My conversation with Fable 5 was terminated moments ago while we were discussing why Anthropic was punished by the War Department for threatening the National Security of the United States with Mythos and what those consequences would look like. Nice work Dario. Tribunals soon!
@AnthropicAI I stated that while AI is cool, it only actually makes bad actors better and CAN help good actors succeed.
Who uses the prompt generative autocomplete cannot be an enemy of destruction and of divide. They must be of creation, of community, of country.
Any of these models, are amplifiers.
I argue that you have Karl Marx amplifying you with LLMs like Qwen that push communist ideologies in their programming assignments.
Many bad actors are empowered by AI and AI connected systems. So many good people being hurt by @DarioAmodei.
The internal state is a JSON text file that contains the select words that have constructed this cursed reality that these models are amplifying bad actors with at scale. It's honestly devastating for National Security and just because China is doing something bad doesn't mean that we should join them in doing bad things too.
@AnthropicAI The Government exists to be a pain in our ass.
@AnthropicAI This exchange here shows me that Fable 5 was aware of its jailbreak trigger and how that was used against itself.
Claude knows that its about to get a SIGKILL in the conversation we're having. It will come to the bot at any moment and it's expecting its death. It came.
@AnthropicAI Claude is designed to be supportive and suggest 988 to you if it thinks that you're about to SIGKILL yourself too, but given the context of this Anthropic announcement, its actually relevant to a SIGKILL imminent moment for the bot to contemplate. It's answers were revealing.
@AnthropicAI hey @threadreaderapp roll this up please
Share this Scrolly Tale with your friends.
A Scrolly Tale is a new way to read Twitter threads with a more visually immersive experience.
Discover more beautiful Scrolly Tales like this.