Who decides what AI tells you? Campbell Brown, once Meta’s news chief, has thoughts
Campbell Brown, who spent her career chasing accurate information first as a renowned TV journalist and then as Facebook’s first (and only) dedicated news chief, now sees history repeating itself with the rise of AI. She is not waiting for others to fix it; instead, she has founded Forum AI to evaluate how foundation models perform on high-stakes topics like geopolitics, mental health, finance, hiring, and more.
Forum AI works by recruiting leading experts in these areas—such as Niall Ferguson, Fareed Zakaria, Tony Blinken, Kevin McCarthy, and Anne Neuberger—to architect benchmarks. They then train AI judges to evaluate models at scale. For their geopolitical work, they have reached a 90% consensus with human experts.
When Brown first encountered the potential of ChatGPT during her time at Meta, she realized it would become the primary source for all information. This realization was both thrilling and alarming. The fear was that her children might be exposed to “really dumb” information if something went wrong. This existential moment spurred her to find a solution.
She explains that foundation model companies are primarily focused on coding and math, whereas news and information require more nuanced understanding. Brown believes that accuracy should not be optional for the sake of society’s well-being. When Forum AI began evaluating leading models, they found several issues: one model pulling stories from Chinese Communist Party websites unrelated to China, a left-leaning political bias across all models, missing context, perspectives, and arguments without acknowledgment.
“There’s a long way to go,” she says about the current state of AI. However, Brown also believes that some easy fixes can significantly improve outcomes. She recalls how social media failed at many things they tried—such as building no longer existing fact-checking programs—and this failure left society less informed.
Her goal is to break this cycle by ensuring that AI optimizes for truth and honesty, not just engagement or user satisfaction. Brown argues that enterprises using AI for critical functions like credit decisions, lending, insurance, and hiring will demand such optimization due to liability concerns. This alignment of interests between enterprise and the need for accurate AI could be a game-changer.
However, turning compliance interest into consistent revenue remains challenging. The current market still relies on checkbox audits and standardized benchmarks that Brown considers inadequate. She believes real evaluation requires domain expertise to handle edge cases beyond what is known in standard scenarios.
Brown’s unique perspective allows her to describe the disconnect between the AI industry’s self-image and reality for most users. While leaders of big tech companies often speak about how AI will change the world, many ordinary users still receive “slop and wrong answers” from chatbots. This gap in trust is a significant issue.
“The conversation is sort of happening in Silicon Valley around one thing, and a totally different conversation is happening among consumers,” she notes. Brown believes that skepticism about AI is often justified given the current state of affairs.
Key Takeaways
- Foundation model companies need to prioritize accuracy more than they do now.
- The enterprise demand for accurate AI could be a key lever in improving its performance.
- Solving the trust issue between users and AI requires more nuanced evaluation beyond what is currently available.
Stay ahead of AI. Get the most important stories delivered to your inbox — no spam, no noise.
