What does the future of AI look like and could it destroy humanity?

A former Anthropic researcher’s resignation thread warning about racing toward self-improving superintelligence, and debates what a realistic “middle ground” on AI safety and regulation could look like.
Google AI mapping a fruit fly brain (166,000 neurons, 125 million connections) and what it could mean for neuroscience, plus a tour through AI’s weird new benchmarks, from Doom and Minecraft fly-brain demos to Blender’s reboot of the classic “Will Smith eating spaghetti.”