Lots of interesting stuff this week, including Google’s new AI Mode, Meta’s superintelligence talent spree, and Musk’s Grok chatbot controversy. Also Perplexity’s Comet is out, but only for big spenders.
But first... a big shout out to our Patron of the Week: Steven Kimbrough! If you want to support the show, check out patreon.com/aiinsideshow.
Google AI Mode Arrives
Jeff finally got AI Mode! Google timed its announcement with Samsung Unpacked, revealing that AI Mode is now integrated into Circle to Search on Android. Just tap AI Inside for Wednesday, July 2, 2025and hold, circle the thing, and jump right into AI mode about that thing. AI Mode also appears inside Google Lens in the Google app, so it touches both Android and iOS.
Meta’s Superintelligence Talent Spree
Meta appears to be on a buying spree as it staffs its superintelligence effort. Apple is losing Ruoming Pang, head of its foundation models team, lured by Meta with "tens of millions of dollars... PER YEAR." Pang managed a team of 100 engineers at Apple working on core AI models. While Apple is seen as playing catchup, Pang’s short tenure at Apple makes the loss hard to gauge. Apple is reportedly considering using models from OpenAI or Anthropic to bulk up its own efforts.
Safe Superintelligence Shakeup
Meta also nabbed Daniel Gross, who had been leading Safe Superintelligence (SSI), the startup co-founded with Ilya Sutskever. Sutskever will now assume the CEO role, while Meta’s acquisition offer was declined. Daniel Levy will serve as President. The company’s mission: build safe superintelligence—without distractions from acquisition offers.
Grok’s Controversial Weekend
Elon Musk updated xAI’s chatbot Grok with new prompts: treat media-sourced views as biased and don’t avoid “politically incorrect” statements. Grok made a series of controversial posts, including blaming Musk and Trump for Texas flood casualties and repeating antisemitic stereotypes. No word on whether the outbursts were the result or the cause of the new prompts.
Meanwhile, Linda Yaccarino, CEO of X, announced her departure, reportedly in the works before the Grok incident.
Generative AI Impersonation Attempt
An impersonator used generative text and voice models to mimic US Secretary of State Marco Rubio, contacting high-level officials by text, Signal, and voicemail. The fake account mimicked Rubio’s voice and text style to access sensitive info. The attempts failed, and the State Department is investigating. This incident confirms fears about AI-driven impersonations, but the good news: it didn’t work.
OpenAI Tightens Security
OpenAI is strengthening security protocols to guard against corporate espionage and IP theft. This follows DeepSeek’s January release, which OpenAI alleges was built using “distillation” techniques. New measures include limiting staff access to sensitive projects, storing sensitive projects offline, mandatory biometric scans, and a deny-by-default internet policy. These steps are standard for protecting trade secrets, especially with ongoing poaching risks.
AI in Advertising: Real-World Examples
Over the weekend, we published our interview with Rishad Tobaccowala about the future of advertising in an AI-driven world.
Rishad predicted generative AI would soon be used to create consumer-facing materials. Two quick examples:
Channel 4’s AI-Generated Adverts:
The UK’s Channel 4 is introducing AI-generated ads on its streaming platform, powered by Streamr.ai and Telana. The goal: make TV advertising accessible to small and medium businesses. The system quickly creates ads from a company’s website and social media, with a broader rollout planned later this year.
Clorox Embraces GenAI:
Clorox, owner of Hidden Valley Ranch, is using genAI to streamline ad creation and test food ads. They’re leaning further into AI to augment, not replace, human creativity and productivity throughout the company.
Perplexity Launches Max Plan and Comet Browser
Everyone’s got a $200 per month plan, right? Now Perplexity has its own! Perplexity Max offers unlimited access to its most advanced tools and priority use of the latest models. Plus, Perplexity officially launched its Comet browser, available first to Max subscribers, with Perplexity search as default and the new Comet Assistant AI agent.
Babak Hodjat on LLMs and AI Understanding
Babak Hodjat, CTO for AI at Cognizant and co-founder of Dejima Inc (the foundation for Siri’s conversational capabilities), discusses how large language models may develop their own abstract ways of representing knowledge. He looks at the challenges and cultural nuances in teaching AI to understand complex concepts like emotion and poetry. The full interview will appear in the podcast feed this weekend!
HUGE Thank You to Executive Producers
DrDew, Jeffrey Marraccini, WPVM 103.7 in Asheville NC, Dante St James, Bono De Rick, Jason Neiffer, Jason Brady, and Anthony Downs!!
See you this Saturday for a special interview with Babak Hodjat