AI assistants Dots and Muse show promise in complex tasks

In a test of OpenAI's Dots and Meta's Muse, a journalist evaluated their capabilities using a WNBA Finals scenario. The task involved planning a fantasy trip to Atlanta, organizing a purple-themed brunch watch party, and creating souvenir programs. Muse is free with optional $20 and $100 monthly subscriptions, while Dots requires a ChatGPT Pro subscription starting at $100 monthly. Both assistants excelled at complex, multistep assignments but struggled with purchasing Valkyries merchandise, settling on WNBA trading cards. They created a 1980s-style souvenir program and transformed Game 1 photos into a keepsake. The journalist hesitated to grant greater access to accounts and wallets, citing trust concerns. The ideal role for these agents is to handle research and legwork, leaving final decisions to the user. The agents showed creativity but were imperfect, capable of turning broad instructions into elaborate products. The journalist emphasized the need for human oversight in consequential decisions, highlighting the gap between AI capabilities and user comfort. The test underscored the ongoing race among AI companies to develop digital assistants for everyday tasks, while revealing remaining limitations and trust issues.