← Feed

Posts

• Post #2480219

  @LoseFriendsandAlienatePeople @nixCraft Very possible and already happening. It’s training data poisoning. https://atlas.mitre.org/techniques/AML.T0020 The attack surface for these models is eye watering. And new vectors are being developed

• Post #2480218

  @LoseFriendsandAlienatePeople This is likely already happening. As governments push biased content through news outlets or directly, future LLMs trained on this data will inherit the same bias.

• Post #2480217

  @neil My t-shirts have hippopotamuses on a triangle’s hypotenuse and a “hippotenuse” caption, or a t-rex holding a cup of tea and a bold “TEA-REX” font. And others similarly unsettling, given I am in my late 40s. So people don’t invite me back. And I win.

• Post #2480216

  @SecurityWriter Maybe we should start calling medicine “oncology” because cancer research gets the funding. So that person still went to the conference? That’s really motivated ignorance. Anyway. It doesn’t surprise me. LinkedIn is nothing more than a status signalling machine. And “I work in AI” sounds better at dinner parties than “I manage network infrastructure”.