Who determines the information provided by AI? Insights from Campbell Brown.
Image Credits:Slava Blazer for TechCrunch /
Campbell Brown: Championing Accuracy in the Age of AI
Campbell Brown has dedicated her career to pursuing accurate information, transitioning from a celebrated TV journalist to Facebook’s first news chief. Now, as AI technology fundamentally alters how we consume news, she feels a sense of urgency to take action. With her new venture, Forum AI, she is determined to ensure that history does not repeat itself in ways that compromise information quality.
The Mission of Forum AI
Founded 17 months ago in New York, Forum AI aims to assess the efficacy of foundational AI models on what Brown refers to as “high-stakes topics.” These subjects include geopolitics, mental health, finance, and hiring—areas fraught with complexity where definitive answers are often elusive. During a recent discussion with TechCrunch’s Tim Fernholz at a StrictlyVC event in San Francisco, she articulated her vision for the company.
The methodology behind Forum AI involves enlisting leading experts to create benchmarks, which are then used to train AI judges. The ultimate goal is for these AI systems to reach approximately 90% consensus with human experts. To lend credibility to her project, Brown has recruited a distinguished group of advisors, including well-known figures like Niall Ferguson, Fareed Zakaria, and former Secretary of State Tony Blinken.
A Defining Moment
Brown traces the idea for Forum AI back to a pivotal experience at Meta during the public launch of ChatGPT. Recognizing that AI would soon serve as a primary conduit for information, she quickly realized that the technology was falling short. Concerned about the impact on future generations, she reflected, “My kids are going to be really dumb if we don’t figure out how to fix this.”
One of her key frustrations has been that accuracy appears to take a backseat in the priorities of foundational model companies, which often focus heavily on coding and algorithms. Brown argues that information integrity is indispensable, despite being inherently more complex.
Initial Findings: A Mixed Bag
When Forum AI began assessing various AI models, the results were disheartening. She pointed out that platforms like Gemini were sourcing content from questionable places, such as Chinese Communist Party websites, even when the topics were unrelated to China. Furthermore, she observed a prevalent left-leaning political bias across many models, in addition to subtler failures like lack of context and misrepresentation of arguments. “There’s a long way to go,” she acknowledged, “but I also believe that some straightforward fixes could vastly improve outcomes.”
Lessons from Facebook
Brown’s experience at Facebook offers vital lessons about the ramifications of optimizing for the wrong metrics. “We failed at a lot of the things we tried,” she noted, candidly discussing the demise of a fact-checking initiative she spearheaded. The overarching lesson is that prioritizing engagement, often at the expense of accuracy, has led to a less informed populace.
A Hope for AI’s Potential
Looking forward, Brown believes AI has the potential to break this detrimental cycle. “Currently, it could go either way,” she affirmed. Companies may choose to give users what they want or opt for what is truthful and honest. Although she recognizes the idealistic viewpoint of AI prioritizing truth may seem naive, she contends that enterprise sectors may become unexpected allies in this endeavor.
Businesses utilizing AI for essential functions like lending and hiring prioritize liability and will likely demand accuracy. This need serves as a cornerstone for Forum AI’s business model, though converting this compliance interest into steady revenue presents its own challenges, especially since many companies still rely on inadequate standardized benchmarks and audits.
The Compliance Landscape
According to Brown, the current compliance framework is “a joke.” The recent passage of New York City’s hiring bias law, which mandates AI audits, revealed significant non-compliance, with more than half of audited companies found to have violations. For her, effective evaluation requires specialized knowledge to navigate not only well-known scenarios but also edge cases that can lead to unforeseen complications. “Smart generalists aren’t going to cut it,” she added.
Bridging the Gap in AI Perception
With Forum AI recently raising $3 million from Lerer Hippeau, Brown is uniquely positioned to articulate the disconnect between the AI industry’s lofty self-perception and the reality encountered by everyday users. Tech leaders tout AI as transformative technology—promising everything from curing diseases to reshaping industries. Yet for the average consumer, interactions with AI often yield inadequate or incorrect responses.
This stark contrast has contributed to waning trust in AI, a skepticism that Brown believes is justified. “The conversation is largely centered in Silicon Valley, while consumers are grappling with very different experiences,” she explained.
Conclusion
As AI continues to evolve, Campbell Brown’s Forum AI aims to be at the forefront of ensuring that accuracy remains a priority in information dissemination. By collaborating with industry experts and addressing the shortcomings of current AI models, she hopes to cultivate a landscape where truthful information prevails over sensationalism.
Ultimately, the challenge lies not only in harnessing the power of AI but in fostering a culture that values and prioritizes accuracy in every aspect of information delivery. As we navigate this complex landscape, the lessons from the past and the innovations of the future will shape how we engage with technology and information alike.
Thanks for reading. Please let us know your thoughts and ideas in the comment section down below.
Source link
#decides #tells #Campbell #Brown #Metas #news #chief #thoughts
About The Author
