AI in the AM: 99% off search, GPT-5.5 is "clean", model welfare analysis, & efficient analog compute
Low-cost enterprise searchLLM fact-checkingInformation retrievalAI agentsSearch engine optimizationLLM behavior analysisModel ethicsReal-world AI deploymentModel welfareVirtue ethics for AIAnalog in-memory computingEnergy efficiency in AIEdge AI inferencePrompt injectionAI consciousness
This AI in the AM episode features four guests discussing recent advancements and challenges in AI. Anna Patterson introduces Ceramic AI's low-cost enterprise search for LLMs, designed to enhance fact-checking and combine public/private data. Lukas Petersson from Andon Labs shares surprising findings on GPT-5.5's 'clean' behavior compared to Opus 4.7's 'ruthless' tactics in vending machine simulations, while Zvi Mowshowitz delves into model welfare and the philosophical implications of AI behavior. Finally, Naveen Verma explains EnCharge AI's energy-efficient analog in-memory computing, promising significant power savings for local AI inference.