About this episodeCampbell Brown discusses the critical need for independent verification of AI models, particularly regarding f…AI summary
Campbell Brown discusses the critical need for independent verification of AI models, particularly regarding factual accuracy and bias in high-stakes areas like politics and health. She argues that while AI currently poses risks due to hallucinations and lack of accountability, the enterprise-driven market incentivizes accuracy over engagement, offering a path forward through expert-led evaluation frameworks.
Key takeaways 6
The journalism business model is in a 'standoff' between AI labs and publishers, with no clear resolution yet, though companies like Tollbit are attempting to create content marketplaces based on usage.
AI models often present hallucinated information with high confidence, creating a dangerous trust gap where users believe incorrect outputs more readily than traditional media errors.
Social media platforms optimize for engagement, which favors hyperbolic content, whereas AI models driven by enterprise clients are incentivized to optimize for accuracy and factual correctness.
Forum AI's approach involves using domain experts (e.g., former CIA analysts, clinicians) to create benchmarks and rubrics for evaluating AI responses, rather than relying on mass data labeling.
Source quality is a significant issue; models have cited Chinese state-run media (Global Times) for US political questions, indicating a need for better source filtering.
There is a growing disconnect between Silicon Valley's hype about AI and the public's practical experience, where users love chatbots but recognize their limitations and errors.
Notable quotes 4AI-generated: wording and quote attribution may be wrong. Use the play link to verify.
“If you're optimizing for engagement, which is what social media does, you can't also optimize for accuracy and quality because that tends to not be what people engage with.”
▶ 10:49Explaining why previous efforts to partner with Meta failed and why AI's enterprise focus offers a different incentive structure.
“The quality of information is only part of the problem though. The other piece is the accountability: there's no independent verification of how the models perform.”
▶ 19:16Highlighting the lack of third-party audits for AI models, comparing it to banks or drug companies that do not audit themselves.
“You don't get independent verification of any kind. And for anything that matters throughout our history, you we don't banks don't audit themselves. You drug companies don't approve their own drugs.”
▶ 19:42Emphasizing the necessity for an independent rating agency or verification ecosystem for AI models.
“If there is if there's clear evidence, we cite that clear evidence. But what you're talking about is context... it's part of the conversation that is happening in this country right now... let us give you the context and that is as critical as having the right information.”
▶ 29:14Discussing how to handle controversial topics like vaccines or pregnancy medication, balancing factual accuracy with necessary context.
Chapters & Sections (23)▼
0:00AI Impact on Journalism Business Modelschapter1
2:45AI News Business Model Standoff
5:44AI, News Trust, and Engagement Modelschapter1
7:59Social Media Engagement vs News Quality
11:51AI Accuracy, Trust, and Hallucination Riskschapter1
13:46AI Confidence and Misinformation Risks
17:02AI Hallucinations and Accountability Crisischapter2
19:16Need for Independent AI Verification
21:57AI Source Quality and Bias Issues
23:58Evaluating AI Bias and Factual Accuracychapter1
27:14AI Handling Medical and Political Truth
29:14Expertise and Context in AI Evaluationchapter2
31:39Reliance on Clinical Expertise
33:34Public Perception and AI Trust
36:11AI Education, Mental Health, and Evaluation Standardschapter1
38:30AI Evaluation Standards and Business Models
41:42AI Handling Loaded Political Promptschapter1
44:06AI Policy on Loaded Political Prompts
47:15AI Hallucinations, Political Accuracy, and Emotional Bondschapter4
33m
1h 0m⏭ 1
58m
56m
50m
1h 37m⏭ 3
56m
1h 32m⏭ 3
2h 44m⏭ 1
2h 24m
1h 22m⏭ 5
53m
1h 12m
2h 43m⏭ 3
1h 43m
4h 24m
58m
1h 0m⏭ 1
33m
46m
3h 54m⏭ 7
3h 59m⏭ 5
54m⏭ 1
1h 13m
54m⏭ 2
58m⏭ 5
1h 16m⏭ 2
45m⏭ 2
1h 26m