EP 160 : WWDC AI Snooze, Music Deals Shakeup, AI Masters Emotional Tests

Today, we're diving into the latest and most impactful developments shaping the world of Artificial Intelligence. We'll cover the expected AI letdown at Apple's upcoming WWDC, significant licensing deals brewing between record giants and music AI startups, and fascinating research showing AI actually beating humans on emotional intelligence tests.Tune in for Today's New Updates where we explore why Apple's WWDC might feel like a "gap year" for AI, potentially pushing bigger reveals to 2026. We discuss how Apple might open its 3B parameter models to developers for custom features, rebrand existing features as "AI-powered," and introduce a new OS naming system to shift perception. However, major projects like an LLM-powered Siri overhaul, health-focused Project Mulberry, and a ChatGPT competitor reportedly remain in limbo. Accuracy concerns and executive disagreements are keeping Apple's 150B model, said to be nearing ChatGPT quality in testing, under wraps. This places Apple in a difficult position, facing high expectations and fast competition with reported limited progress on the AI front.In the music world, major labels Universal, Warner, and Sony are reportedly in talks with AI music leaders Udio and Suno for licensing deals. These negotiations aim to resolve billion-dollar copyright infringement lawsuits filed in 2024, potentially seeking up to $150k per infringed work. A deal would end the lawsuits and establish a framework for compensating artists whose work is used in training, with labels reportedly seeking licensing fees and equity stakes. This mirrors past industry conflicts but shows labels leaning towards negotiation over prolonged battles, seeing AI licensing as a potentially lucrative revenue stream.We also look at new AI research, including findings from the University of Geneva and University of Bern that show AI systems like ChatGPT outperforming humans on standard emotional intelligence assessments. Six tested AI models averaged an 81% score on selecting emotionally appropriate responses to complex scenarios, compared to just 56% for human participants. GPT-4 even proved capable of creating entirely new emotional intelligence tests. Researchers believe this indicates AI's grasp of emotional concepts and reasoning, not just pattern regurgitation, highlighting potential for AI integration in mental health support, customer service, and education.Plus, get updates on ElevenLabs' Conversational AI 2.0, enabling the creation of custom voice agents with low latency and automatic language switching; Sakana AI's self-improving Darwin Gödel Machine, an LLM capable of rewriting its own code to boost performance; and Meta's AI-powered headsets for soldiers developed with Palmer Luckey's Anduril, designed to sharpen senses for detecting drones or hidden enemies. Learn about Anthropic's new open-sourced framework for probing AI's "internal reasoning steps" and understanding models' thought processes, including the finding that Claude thinks in a "universal language".Stay tuned for Everything else in AI today, featuring news like OpenAI's plans for "ambient" hardware devices, Anthropic's reported $3B in annualized revenue, Meta automating privacy risk assessments with AI, the scale of video generation with Google DeepMind's Veo 3, Google's AI Edge Gallery app, Abridge AI's significant funding round, the public availability of Gemini Live and the Jules coding assistant, Midjourney's latest speed and moderation updates, and the report showing China is now neck-and-neck with the US in open-source AI development.Finally, we'll highlight some Today's Trending AI Tools like Kontext for image editing, Perplexity Labs for building reports, Manus Slides for presentations, EVI 3 for custom voices, Zero for email management, Ciro for prospect list building, and Pretty Prompt for refining AI prompts.Don't miss this packed episode of the AI Brief!