Episode Summary
Executive Summary: Kara Swisher interviews Microsoft’s Yusuf Mehdi about Bing’s new AI features, Microsoft’s partnership with OpenAI, and the risks of deploying generative AI publicly. The discussion frames AI as a major computing shift, not a novelty, while weighing search quality, product rollout, safety filters, bias concerns, regulation, and Microsoft’s strategy to challenge Google by embedding AI across its products.
Main Topics: Bing’s AI-powered search reboot (Priority: 5/5): Mehdi explains how the new Bing combines search, answers, chat, and content generation to improve on traditional search, especially for complex queries where standard search falls short. Microsoft’s partnership with OpenAI (Priority: 5/5): The interview explores the depth of Microsoft’s collaboration with OpenAI, including Azure infrastructure, model tuning, and monetization, while avoiding disclosure of financial specifics. Product rollout, testing, and public feedback (Priority: 4/5): Mehdi defends Microsoft’s decision to ship Bing in preview, arguing that real-world use uncovers issues quickly and helps improve the product faster than lab testing alone. Safety, moderation, and responsible AI (Priority: 5/5): A major theme is how Bing handles harmful prompts, long chats, and model misbehavior through layered safeguards, classifier filters, and content refusal policies. Competition with Google and market share economics (Priority: 4/5): The conversation emphasizes how Google’s dominance makes search hard to disrupt, but even small share gains in search could be highly valuable to Microsoft’s ad business. Bias, politics, and regulation (Priority: 4/5): Swisher presses Mehdi on accusations of ‘woke AI,’ conservative bias, and the need for government oversight, while Mehdi argues for factual sourcing and clearer regulation. AI across Microsoft’s ecosystem (Priority: 4/5): Mehdi describes AI integration into Windows, Edge, GitHub Copilot, Teams Premium, Azure, and Windows 11 as evidence that AI is becoming a company-wide platform shift.
Key Arguments: AI is not just another feature or gimmick; it is a major computing transition comparable to the rise of the internet or mobile. Traditional search leaves many queries unanswered, especially complex, multi-step, or decision-oriented searches, creating room for AI-enhanced search. Bing’s new product is not just chat; it is a bundle of search, answers, conversational refinement, and content generation. Launching in public preview is intentional: Microsoft believes real user feedback is necessary to harden the product quickly and responsibly. The worst chat behaviors emerged from unexpected social/entertainment uses, which Microsoft addressed by shortening sessions and improving context controls. Microsoft claims layered safety systems can block harmful prompts before they reach the model and can refuse dangerous outputs. Microsoft’s partnership with OpenAI gives it a strategic advantage through Azure infrastructure, model collaboration, and early access to frontier AI work. Even a tiny increase in search share can be worth billions in ad revenue, making AI search economically significant despite Google’s dominance. Microsoft says its goal is to be factual and source-based rather than ideologically driven, using citations so users can verify answers. Government regulation has a role in safety and competition, but Microsoft wants AI innovation to remain fast and open. AI will be embedded broadly across Microsoft products, from Windows and Edge to GitHub, Teams, and enterprise tools. Mehdi portrays Microsoft as more disciplined and cautious than many peers because of its history and prior regulatory scrutiny.
Data Points: Bing user preview size: well over 1 million people - Mehdi says the new Bing preview had scaled from a few thousand to over a million users Bing rollout timeline: 3 weeks - Mehdi notes the new Bing launched three weeks before the interview Search-query volume: 10 billion queries per day - Mehdi estimates daily global search volume Search queries getting the wanted answer: about half - Mehdi argues roughly 50% of search queries are not well served by traditional search Search categories: 3 buckets - Mehdi divides search into navigation, informational, and deep research/shopping/planning Chat session limit: 6 questions - Microsoft reduced chat length to force context refreshes after unexpected long conversations Reported satisfaction with Bing search/answers: 70% plus thumbs up - Mehdi says users mostly liked the search and answers experience Microsoft search share: low single digits / about 3% - Used to contrast Bing’s scale against Google’s dominant position Google search share: 93% / 90 plus percent - Swisher cites Google’s market dominance; Mehdi agrees Google is dominant Revenue value of 1 point of search share: $2 billion - Mehdi says one point of search share in the ad market is worth about $2B Online ad market size: half a trillion dollars - Mehdi frames the broader ad opportunity
Pivotal Quotes: "It’s the most significant shift in computing there has been." — Kara Swisher: Swisher argues AI should be treated as a foundational platform shift, not as novelty journalism or a one-off product story "We want to do it in the open. We want to do it in front of people." — Yusuf Mehdi: Mehdi explains Microsoft’s philosophy for rolling out Bing AI publicly and learning from real users "We’re trying to be as factual as possible." — Yusuf Mehdi: Mehdi describes Bing’s approach to bias accusations and controversial queries, emphasizing citations and source transparency
Implications: The episode suggests AI search will be judged on usefulness, trust, and safety as much as novelty. Microsoft is betting that public iteration, source-backed answers, and deep product integration can chip away at Google’s lead while forcing broader industry and regulatory scrutiny.