Is Google Handing AI Startups the Keys?
AI Inside for Wednesday, September 3, 2025
Hey there everyone! We recorded a new episode of the AI Inside podcast (from my garage because the bathroom remodel is getting NOISY!) Some pretty major news stories, including Google’s new legal reality, Netflix’s data-driven movie revolution, and Tesla’s big robot gamble.
But first… let’s call out our Patron of the Week: Burke Norton and shout out to Steve Remington for upgrading! You can support us at patreon.com/aiinsideshow.
Google’s Legal Shakeup and the AI Search Race
US Judge Amit Mehta ruled against some of the most punishing penalties requested in the DOJ’s search monopoly case. Google keeps Chrome, Android, and its Apple search deal, but exclusivity clauses are out. But here’s the thing: Google must share parts of its search data with qualified competitors, setting up a potential playing field for AI companies like Perplexity and OpenAI. Judge Mehta even noted GenAI companies are now better positioned competitively. Could this finally give up-and-comers a chance to rival Google Search? Google is of course considering its appeal.
The Superintelligence “Fever Dream” Breaks
Gary Marcus wrote in the New York Times: the “fever dream of imminent Superintelligence is finally breaking.” GPT-5 has disappointed experts by not moving past old issues and hallucinations, aka no sign of AGI here. Marcus says we need more cognitive science, world models, and symbolic reasoning, plus actual legislative checks.
Meanwhile, Meta’s superintelligence “TBD Lab” remains chaotic, with top hires leaving before onboarding. Reports say Meta won’t release its big Llama Behemoth model due to lackluster results.
ChatGPT’s Mental Health Safeguards
Friend Megan Morrone reports on new changes coming to ChatGPT: renewed commitment to protecting teens and people in distress. OpenAI will roll out guardrails by year’s end, with parental notifications, the ability to disable history, memory, and escalate concerning threads to human review. Therapists warn of emotional dependency on chatbots and self-diagnosis pitfalls. Over in China, lonely and sick users turn to chatbots for support and medical info, showing both the promise and risks.
Netflix, Data, and the Algorithm Movie
The “algorithm movie” is real, especially at Netflix. Data-driven films get greenlit to maximize reach, but insiders admit the results can be generic and forgettable. AI models tempt execs to keep chasing what data says will hit, often at the expense of originality. Netflix alone collected 700 billion data events in 2017.
The New Yorker points out that AI is already changing how we make and consume culture and that there’s opportunity for creativity, but also a risk of cultural erosion.
Tesla’s Robot Dreams
On X, Elon Musk said Optimus, Tesla’s humanoid robot, will represent 80% of Tesla’s future value, not the cars. He wants 5,000 robots built this year and sees robots, plus robotaxis, as Tesla's real future. That’s ambitious, especially as car sales slump and Tesla just lost its robotics division head.
Amazon Lens Live and More AI in Everyday Life
Amazon launched Lens Live on iOS. AI lets users shop by simply aiming the camera at real-life items and products are instantly matched, with summaries and recommendations powered by AI assistant Rufus. Android support is coming soon.
AI Stethoscopes: Real Impact
Doctors at Imperial College London created an AI-powered stethoscope that can diagnose multiple heart conditions in just 15 seconds. It’s doubled heart failure detection and tripled Afib detection in trials with 12,000 patients, offering rapid, accurate results sent straight to a phone.
OpenAI’s gpt-realtime Opens the Voice Agent Floodgates
OpenAI’s gpt-realtime model and API now let developers build advanced voice agents with improved voice quality, live phone support, and tight instruction following. This means next-gen AI-driven customer support agents could now be handled by software alone, with advanced permissions baked in.
WordPress Telex: AI for Site Builders
WordPress revealed Telex, an experimental AI tool. It lets users type out a request, like “create a testimonial carousel”, and the AI builds a ready-to-use content block, exported as a simple .zip plugin. Early days, but a big glimpse of what’s coming for creators.
Anthropic Claude’s Chrome Agent
Anthropic’s Claude now has a Chrome extension (for Max subscribers) that creates a context-aware sidecar, lets users chat, and can automate browser tasks, with clear permission prompts to keep things safe. Anthropic’s pushing for reduced vulnerabilities and sensible safety limits. There’s also a public waitlist.
Microsoft’s Own AI: MAI-Voice-1 and MAI-1-Preview
Microsoft AI launched two in-house models: MAI-Voice-1 (already in Copilot), which focuses on expressive speech for bots and virtual agents, and MAI-1-Preview for summarization and insights with both trained with Microsoft’s infrastructure, easing reliance on outside partners.
HUGE thank you to Executive Producers on the Patreon:
DrDew, Jeffrey Marraccini, Radio Asheville 103.7, Dante St James, Bono De Rick, Jason Neiffer, Jason Brady, Anthony Downs, and Mark Starcher!!
See you next Wednesday on another episode of AI Inside.