AI models may be accidentally (and secretly) learning each other’s bad behaviors

4 weeks ago 8

Artificial intelligence models can secretly transmit dangerous inclinations to one another like a contagion, a recent study found