AI developers test ingenuity with Pictionary and Minecraft games

AI developers are exploring new ways to test artificial intelligence by using games like Pictionary and Minecraft. Paul Calcraft has created an app where AI models play a Pictionary-like game, challenging them to think creatively rather than rely on memorized answers. Similarly, 16-year-old Adonis Singh has developed a tool called Mcbench that allows AI to control a character in Minecraft, testing its problem-solving skills in a more open environment. Both projects aim to provide benchmarks that are harder to "game" than traditional tests. While using games for AI testing is not new, the focus on large language models offers a fresh perspective. Researchers believe these games can reveal insights into AI reasoning and decision-making, although opinions vary on their effectiveness compared to other benchmarks.