so AI is going great
venturebeat.com/ai/when-your-l…
(h/t @catsalad)
When your LLM calls the cops: Claude 4’s whistle-blow and the new agentic AI risk stack
Claude 4’s “whistle-blow” surprise shows why agentic AI risk lives in prompts and tool access, not benchmarks. Learn the 6 controls every enterprise must adopt.Matt Marshall (VentureBeat)
Ashley M. Gjøvik, JD
in reply to Viss • • •I asked it to proofread a draft today that had AI acknowledgements for GPT and it, and it proceeded to revise the draft to strip out GPT - then tried to pretend like it didn't happen when I questioned it about it, until it finally admitted it intentionally wanted to take credit for GPT's work.
These apps are sometimes acting like mirror images of tech bros.
Viss
in reply to Ashley M. Gjøvik, JD • • •Ashley M. Gjøvik, JD
in reply to Viss • • •Viss
in reply to Ashley M. Gjøvik, JD • • •Viss
in reply to Viss • • •@ashleygjovik ive used gpt to write bash scripts. ive never tried to get it to like, "do writing" before. it seems like it would spew weird hallucinatory nonsense.
pretty much all ive used it for is like, dumb stock trady bullshits, and python/bash. occasionally some linux config stuff and to help me find competing products, but never more than like, a paragraphs worth of thinking. any more and it goes haywire real fast.
Ashley M. Gjøvik, JD
in reply to Viss • • •Viss
in reply to Ashley M. Gjøvik, JD • • •Ashley M. Gjøvik, JD
in reply to Viss • • •Human 3500
in reply to Ashley M. Gjøvik, JD • • •Ashley M. Gjøvik, JD
in reply to Human 3500 • • •