Huburistic Superintelligence

AI Inside for Wednesday, June 11, 2025

Lot's of interesting stuff this week, including Apple’s bold research paper challenging AI “reasoning,” Meta’s superintelligence moonshot, and the creative collision of AI and Hollywood.

But first... let’s call out our Patron of the Week: Mike Reiser! If you want to support the show, head over to patreon.com/aiinsideshow.

Apple’s Research Paper: “The Illusion of Thinking”

Apple published a research paper titled "The Illusion of Thinking: Understanding the Strength and Limitations of Reasoning Models via the Lens of Problem Complexity." It challenges industry claims about advanced reasoning models, arguing they don’t actually “think”, they just create illusions of reasoning. Apple tested OpenAI’s o3, Claude 3.7, and Google’s Gemini, noting a “complete accuracy collapse” on complex problems, even with enough tokens. Steven Sinofsky praised the paper, calling out the “hubris” in AI. Nate Jones critiqued the study for its limitations, like not giving models tools or internet access, and suggested models need a framework to “call for help” when they hit their limits. Lightweight models handle 98%, while the most advanced models tackle the toughest 2%.

Ilya Sutskever’s “Digital Brain” Speech

Ilya Sutskever, co-founder and former chief scientist of OpenAI, received an honorary doctorate from the University of Toronto and delivered a speech about AI’s future. He stated that we’re headed toward a time when technology can do everything a human can do: “We have a brain, the brain is a biological computer, so why can't a digital computer, a digital brain, do the same things?” Sutskever assured graduates that AI will be able to learn anything they can, raising questions about the future of human achievement in an AI-driven world.

Meta’s Superintelligence Drive

Meta is launching a new AI research lab focused on superintelligence, with Mark Zuckerberg personally leading recruitment for a team of about 50. The company is reorganizing around this goal and investing $14 billion into Scale AI, whose CEO Alexandr Wang will join Meta in the process. Scale AI has helped major players like OpenAI and Google with training data. Zuckerberg, frustrated with Llama 4’s performance and delays on their “behemoth” model, is now in “Founder Mode,” taking a hands-on approach. The big questions: What is superintelligence? What is AGI? How will we know when we get there?.

Apple’s On-Device AI Strategy

At WWDC, Apple revealed a new Foundation Models framework, enabling apps to add AI features without cloud costs and with privacy in mind. The framework is built into the OS, so apps don’t increase in size and developers avoid extra charges. This on-device approach fits Apple’s ecosystem philosophy, allowing unlimited AI usage with no cloud API fees.

NotebookLM’s Public Sharing and Research Workflow

NotebookLM now lets users share notebooks via a simple link, making them public for viewing and engagement (though not editing). You can generate and listen to audio overviews.

Manoor Faisal at XDA combined NotebookLM with Perplexity for a powerful research workflow: Perplexity excels at real-time web search, while NotebookLM organizes and analyzes the findings. This has me wishing for an automated way to bridge these tools… maybe something like MindStudio?.

Huxe: The AI-Generated Podcast Startup

Three former NotebookLM team members have launched Huxe, an AI startup focused on audio delivery and AI-generated podcasts. Huxe connects to personal data streams like Google Calendar and Gmail, delivering daily audio briefings and “DeepCast” responses to user questions. It’s currently iOS-only, with invite codes available on Twitter. Impressively, the team built Huxe in just one month, a pace that they likely couldn’t achieve working for the Google mothership.

AI and Creativity: Primordial Soup and “Artificial”

Google DeepMind’s Demis Hassabis and filmmaker Darren Aronofsky partnered on Primordial Soup, producing “Ancestra”—a hybrid live-action/AI short that premiered at Tribeca. Both believe AI is useful for many things, but true storytelling invention remains a human skill, at least for now.

Meanwhile, Luca Guadagnino’s film “Artificial” dramatizes Sam Altman’s firing and rehiring at OpenAI in 2023. Fast-tracked for Amazon MGM Studios, it’s fascinating to see films made so close to real events, often before there’s enough perspective to fully understand them.

Hollywood’s Quiet Embrace of AI

A recent Vulture article highlights Hollywood’s increasing use of AI tools, even as the industry hesitates to admit it publicly. Workers at every level—concept art, storyboards, visual effects—are experimenting with AI. The article underscores the obvious: these tools are impacting everything, and people are gradually recognizing their value, even if they’re reluctant to say so.

Thank you for watching and reading!

HUGE thank you to Executive Producers on the Patreon: DrDew, Jeffrey Marraccini, WPVM 103.7 in Asheville NC, Dante St James, Bono De Rick, Jason Neiffer, Jason Brady, and Anthony Downs!!

See you next Wednesday on another episode of AI Inside.