Campbell Brown: AI Must Choose Between Truth and Engagement

Campbell Brown and Her Vision for AI in Journalism

Campbell Brown, a recognized figure in the field of journalism, has dedicated much of her career to the pursuit of accurate information. After being a renowned television journalist, she took the helm of news at Facebook, becoming the first person to hold that position. Today, as she observes artificial intelligence transforming the way people access information, she expresses her fears about a repetition of past mistakes. This time, she does not want to rely on others to find solutions.

Forum AI: An Innovative Approach to Evaluating AI

Brown founded Forum AI, a company focused on assessing the performance of AI models in areas she describes as "high-stakes topics." These areas include geopolitics, mental health, finance, and recruitment. These subjects are characterized by their complexity and the absence of simple answers. The goal of Forum AI is to collaborate with the world's leading experts to establish benchmarks, and then train AI judges capable of evaluating models at scale.

For her work on geopolitics, Brown has gathered influential figures such as Niall Ferguson, Fareed Zakaria, former Secretary of State Tony Blinken, former House Speaker Kevin McCarthy, and Anne Neuberger, who led cybersecurity under the Obama administration. The aim is for the AI judges to reach a consensus of around 90% with these human experts, a threshold that Forum AI claims to have already achieved.

The Origins of Forum AI and Its Motivations

Forum AI was established 17 months ago in New York, at a time that Brown describes as crucial. She recalls her time at Meta when ChatGPT was first launched. She quickly realized that this technology would become the primary channel for disseminating information, which deeply concerned her. The implications for her children made this realization almost existential, fearing they would become "really dumb" if a solution was not found.

The Challenges of Accuracy in AI

Brown has been particularly frustrated by the lack of attention given to accuracy in AI development. She criticizes foundational model companies for their excessive focus on coding and mathematics, at the expense of information quality. According to her, while the task is complex, it should not be neglected.

During the initial evaluations at Forum AI, the results from leading models were not encouraging. Brown cited the example of Gemini, which sourced information from Chinese Communist Party websites for topics unrelated to China. She also noted a left-leaning political bias in almost all models. Other failures include a lack of context and perspectives, as well as arguments that are overly simplified.

Lessons Learned from Experience at Facebook

Brown spent several years at Facebook, observing the consequences of poorly directed optimization. She told Tim Fernholz that many initiatives had failed, including the fact-checking program she had implemented. She learned that optimizing for engagement was harmful to society and contributed to increased misinformation.

Hope for a Better Future with AI

Despite the challenges, Brown hopes that AI can break this cycle of misinformation. She believes that companies have a choice between providing users with what they want or offering them truthful and honest information. While the idea of an AI optimized for truth may seem idealistic, she believes that the private sector could play a key role. Companies using AI for critical decisions, such as credit or recruitment, have a vested interest in ensuring accurate outcomes.

The Economic Challenges of Forum AI

Forum AI is banking on this demand for accountability for its business model, although transforming this interest into stable revenue remains a challenge. The current market is often satisfied with superficial compliance audits, which Brown considers insufficient. She criticizes the compliance landscape, calling it a "joke," and emphasizes the need for real expertise to address complex scenarios.

The Gap Between Perception and Reality of AI

With $3 million in funding raised last fall, led by Lerer Hippeau, Brown is well-positioned to observe the gap between the perception of the AI industry and the reality for users. She notes that, despite the grand promises from tech leaders, users often receive incorrect answers from chatbots. Trust in AI is low, and Brown believes this skepticism is often justified. She highlights the difference between discussions in Silicon Valley and the concerns of consumers.