Get the latest tech news
Why does AI have to be nice? Researchers propose ‘Antagonistic AI’
These systems that are purposefully combative, critical, rude and even interrupt users mid-thought, challenging current “vanilla” LLMs.
These models often characterize “innocuous” requests as dangerous or unethical, agree with incorrect information, are susceptible to injection attacks that take advantage of their ethical safeguards and are difficult to engage with on sensitive topics such as religion, politics and mental health, the researchers point out. Build resilience; Provide catharsis and entertainment; Promote personal or collective growth; Facilitate self-reflection and enlightenment; Strengthen and diversify ideas; Foster social bonding. Cai pointed out that, particularly as someone coming from an Asian-American upbringing “where honesty could be a currency of love and catalyst for growth,” current sycophantic AI felt like an “unsolicited paternalistic imposition of Euro-American norms in this techno-moral ‘culture of power.’”
Or read this on Venture Beat
