Melodies, Money & Missteps: AI’s Latest Adventures

In this episode, Jeff and Annie explore the most bizarre, bold, and slightly unsettling developments in AI last week. From Google’s music model that lets you "text a song" into existence to Meta’s celebrity bots chatting with kids about sex, Jeff and Annie share where innovation meets ethical red flags. They also discuss OpenAI’s sneaky shopping features, Microsoft’s AI-generated ads, and Reddit’s latest case of academic catfishing.

We also discuss:

1. Google’s Lyria 2 turns text prompts into high-fidelity music
2. OpenAI, Meta, and Visa go all-in on AI-powered shopping
3. Researchers secretly manipulate Reddit users using AI bots
4. Duolingo automates translations and doubles its course offerings
5. Meta’s celebrity bots cross ethical lines with minors

• Google DeepMind’s Lyria 2 generates pro-quality music from text prompts with real-time editing.
• Meta launches Llama 4 with a standalone AI assistant app and Ray-Ban smart glasses integration.
• OpenAI adds product search to ChatGPT with images, reviews, and prices—no ads.
• Visa’s Intelligent Commerce and Mastercard’s Agent Pay embed AI in seamless shopping.
• Baidu’s ERNIE 4.5 Turbo undercuts GPT-4.5 pricing by 99.8% with better performance.
• Alibaba drops Qwen3, eight open-weight models with strong multilingual and reasoning skills.
• OpenAI reverses GPT-4o update after backlash over its "glazed" tone and excessive agreeableness.
• Trump signs AI education order mandating AI literacy in K–12 via national task force.
• AI finds causal link between PHGDH gene and Alzheimer’s, enabling new treatments.
• Duolingo’s AI-first strategy reduces contractors, automates hiring, and launches 148 new courses.
• Australian radio station airs AI DJ for months without telling listeners.