source: techcrunch ai: who decides what ai tells you? campbell brown, once meta’s news chief, has thoughts
level: business
campbell brown, once meta's news chief, now runs forum ai, a startup that tests how foundation models handle nuanced topics. she focuses on areas like geopolitics, mental health, finance, and hiring, where answers are rarely clear-cut. forum ai recruits top experts to create benchmarks and trains ai judges to match their assessments, aiming for 90% consensus. brown says they have reached that threshold in some domains.
brown's concern stems from her time at meta, where she saw chatgpt's release and realized ai would become the main information funnel. she worried about accuracy, especially for her children. early evaluations by forum ai revealed problems: gemini pulled from chinese communist party sites for unrelated stories, and most models showed left-leaning political bias. other issues include missing context, overlooked perspectives, and straw-man arguments.
brown believes enterprise demand for reliable ai could drive improvement, as businesses need accuracy for credit, lending, and hiring decisions. however, she calls current compliance audits a joke, citing new york city's hiring bias law where over half of audits missed violations. forum ai, which raised $3 million, argues that real evaluation requires domain experts to catch edge cases, not just generalists. brown notes a disconnect between tech leaders' hype and users' experience of slop and wrong answers.
why it matters: evaluating ai on complex, high-stakes topics is crucial for trust and safety, especially as models influence decisions in hiring, finance, and public discourse.
source: techcrunch ai: who decides what ai tells you? campbell brown, once meta’s news chief, has thoughts