The AI Radio Show with Arthi | S02 Ep.3

Deep Dive Wednesday: When AI Starts InventingWelcome back to The AI Radio Show.This Deep Dive Wednesday takes us somewhere special: the moment AI crosses from doing to discovering.In this episode, we crack open Google DeepMind’s AlphaEvolve, an evolutionary coding agent that is not just optimizing code, but inventing entirely new algorithms. We explore how AlphaEvolve teaches itself, tests itself, and improves itself using large language models like Gemini 2.5, forming a self-improving loop that feels less like software and more like a co-scientist.You’ll hear how AlphaEvolve achieved what decades of human effort could not, improving matrix multiplication algorithms that stood unchanged for 56 years, speeding up GPU and TPU performance, optimizing data center scheduling, and unlocking efficiency gains that translate into massive real-world impact at Google scale.Along the way, we check the algorithmic weather report, covering:Multimodal leaps with Gemini 1.5 ProCreative turbulence from Midjourney v6Hardware tailwinds with NvidiaRegulatory clouds forming via the EU AI ActWe pause for a synth-powered sonic reset, then shift gears with a PinkFare spotlight, where AI is used not for scale or speed, but for fairness, exposing hidden price discrimination through data.The episode wraps with a hands-on Try This micro-challenge, inviting you to use AI creatively on everyday problems, followed by an Analog Moment to breathe, ground yourself, and remember the human at the center of all this intelligence.This is not just an episode about a paper.It’s about a turning point.AI is no longer just executing instructions.It’s starting to invent.Stay curious.You’re tuned into the future.🎙️ Want to sponsor the show or collaborate? Email us at: [email protected] in to The AI Radio Show with Arthi—where we decode the signal from the noise This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit thearthiaicollective.substack.com