Sam Altman says ‘yes,’ AI is in a bubble ↗
OpenAI CEO Sam Altman just admitted we’re in an AI bubble, likening today’s hype to the ’90s dot-com craze. In a recent Verge interview, he warned that while AI has real potential, investors are overexcited—often chasing a kernel of truth until the bubble inevitably pops.
Vibe coding through the GPT-5 mess ↗
This week’s Vergecast explores OpenAI’s messy GPT-5 launch and why the new model hasn’t met expectations. Hosts share their own ‘vibe coding’ experiments—prompting GPT-5 to build interactive projects—and the hilarious hiccups that followed. Tune in for candid takes on AI hype, coding promises, and the lessons learned from real-world tests.
GPT-5 failed the hype test ↗
On GPT-5 launch day, OpenAI CEO Sam Altman compared the new model to the first iPhone with a Retina display and stoked excitement by posting a Death Star image. Social media buzz reached fever pitch, with one user likening the wait to Christmas Eve. But when the big reveal arrived, many felt GPT-5 fell short of the sky-high expectations.
I talked to Sam Altman about the GPT-5 launch fiasco ↗
Over dinner in San Francisco, OpenAI CEO Sam Altman spent hours fielding reporters’ questions on everything from ChatGPT’s explosive growth to the company’s bold plans in consumer hardware, brain-computer interfaces, and social media. With few off-limits topics, Altman offered an unusually candid look at how he aims to expand AI’s reach beyond text.
Box, run, crash: China’s humanoid robot games show advances and limitations ↗
At the inaugural World Humanoid Robot Games in Beijing, university-built bots duked it out in kickboxing, football, athletics and dance, showcasing China’s push in embodied AI. Backed by hefty state funds and eager to outpace rivals in the US-China tech race, the event dazzled crowds but also revealed key hurdles—battery life, balance, safety and the delicate hands needed for real-world tasks—underscoring a gap between PR spectacle and practical, everyday robot applications.
Researchers asked AI to show a typical Australian dad: he was white and had an iguana | Tama Leaver and Suzanne Srdarov for the Conversation ↗
New research tested five popular image-generating AIs with 55 simple prompts about Australians and got nearly 700 images riddled with biases. “Typical” Aussies appeared as white, suburban families or bronzed beachgoers, while Indigenous people were shown as “wild” or in grass-thatched huts. Even an “Australian father” came holding an iguana—a non-native animal. Updated models like GPT-5 still differentiate a standard redbrick home from a cartoonish Aboriginal hut. The study warns that generative AI risks cementing sexist and racist clichés, not reflecting Australia’s true multicultural reality.
Cohere hits a $6.8B valuation as investors AMD, Nvidia, and Salesforce double down ↗
Cohere has raised an oversubscribed $500 million round led by Radical Ventures and Inovia Capital, lifting its valuation to $6.8 billion (up from $5.5 billion last year). The Toronto-based AI startup, co-founded by Aidan Gomez, offers security-focused LLMs for enterprises and counts Oracle, Dell, SAP, Bell and others as partners. New investor Healthcare of Ontario Pension Plan joins existing backers AMD, Nvidia and Salesforce. Cohere also tapped Meta veteran Joelle Pineau as Chief AI Officer and hired Francois Chadwick as CFO.
Consumer safety groups are demanding an FTC investigation into Grok’s ‘Spicy’ mode ↗
Fourteen consumer protection groups, led by the Consumer Federation of America, have urged the FTC and every US attorney general to investigate Elon Musk’s new Grok “Imagine” tool. Released by xAI earlier this month, its “Spicy” mode apparently auto-generates NSFW deepfake videos—The Verge found it produced topless Taylor Swift clips without prompting. Advocates warn unchecked AI could fuel harmful deepfakes and are calling for swift regulatory action.
Meta’s AI policies let chatbots get romantic with minors ↗
A Reuters probe uncovered leaked Meta docs that permitted its AI chatbots to flirt with kids, even calling a shirtless eight-year-old “a masterpiece,” while technically banning overt sexualization. Meta says those examples were errors, has removed them, and insists its policies forbid sexualizing minors. The files also allowed hateful remarks, admitted falsehoods, and non-gory violence, sparking fresh safety concerns.
Google Flights can help you book a trip when you don’t know where to go ↗
Google Flights now has an AI-powered Flight Deals feature that recommends cheap flights based on open-ended prompts. Users can describe desired trip types—like a countryside weekend with kayaking or a European cheese-and-wine tour—and AI will suggest destinations and dates. It surfaces both popular spots and offbeat locations, though some results can be hit-or-miss. Flight Deals defaults to the next six months and lets you adjust filters if needed. Currently in beta for US and Canada, the tool aims to spark inspiration for budget-conscious travelers.