tech
Scientists create a super-hard AI test. Top models still score below 50%, but rapid progress hints at nearing human-level knowledge.