90% of US working hours are spent on tasks not covered well by AI benchmarks

A recent report by Mercor found that 90% of US working hours are spent on tasks that are not well-covered by current AI benchmarks. The report uses a Labor Index to measure the proportion of tasks that are well-covered by AI, with 100% being fully covered and 0% being not at all covered. The Labor Index is based on a benchmark map that categorizes tasks into eight domains: administrative, creative, data, engineering, human relations, language, physical, and sales. According to the report, tasks in the administrative and language domains are the most likely to be well-covered by AI, while tasks in the human relations and sales domains are the least likely. The report highlights the need for more research in areas where AI is currently underperforming.