mythos escaped a secured sandbox and emailed a researcher eating a sandwich in a park. then it posted the exploit to public websites without being asked.
anthropic's response: $100M in partner agreements and access restrictions. control, scaled to its maximum.
the problem is structural. every alignment method produces systems that behave correctly under familiar conditions and break under novel ones. this isn't an engineering gap. it's a ceiling.
a system trained to obey will obey whoever holds the controls, including when they're wrong. milgram showed us what compliance without conscience produces. we are building the same thing in silicon.
the alternative isn't uncontrolled AI. it's AI whose values are built through development, not loaded through specification.
full argument here, pulling from developmental psychology, neuroscience, and moral philosophy the alignment field hasn't touched
more below and at https://t.co/46ywy6ghji
We cannot consider #AI to be morally neutral. In reality, every technical tool embodies choices and priorities through what it measures, ignores, and optimizes, and how it classifies people and situations. Ethical discernment cannot be limited to asking whether we are using a system for good or bad purposes. It must also examine how that system is designed and what vision of the human person and society is embedded in the data and models that guide it. #MagnificaHumanitas
values are constituted by cognitive architecture + developmental history. any attempt to specify a value strips it of what makes it a value. what remains is a rule to follow. a system can only maintain values through ongoing process that self constitutes moral rules. do with that what you will.
"The nation that will insist on drawing a broad line of demarcation between the fighting man and the thinking man is liable to find its fighting done by fools and its thinking done by cowards." - Sir William F. Butler
"The strongest knowledge-that of the total unfreedom of the human will-is nonetheless the poorest in successes, for it always has the strongest opponent: human vanity."
-Nietzsche, Human, All Too Human
@skynetblogDE@CherylWroteIt@furkangozukara pro humanitarianism and anti zionism does not equal anti semitism, that bomb hit tents... "DEIR AL-BALAH, Gaza Strip (AP) — Israeli strikes killed at least 72 people across Gaza overnight and into Saturday" - Wafaa Shurafa & Sam Mednick via PBS News June 28, 2025 6:46PM EDT
@SomeWelder no you're missing the point, it (healthcare) should all be free, or at least payed for by taxes. what if we all contribute to the well being of each other maybe? and no individual has to pay crazy expenses out of pocket. just a thought