Today, Alex Tamkin from Anthropic comes over to Booth to talk about Anthropic’s Economic Index. The Booth lecture room is overflowed with enthusiastic audience:
The accompany report covers a lot of insights. What I found interesting:
- Learing-by-doing: skill in collaborating with AI is itself a learnable competency. Users who have been on Claude for 6+ months have a ~4 percentage point higher success rate in their conversations, and this holds up even after controlling for the type of task, country, model choice, and language.
- Learning curve: early adopters with high-skill tasks are pulling ahead, while later, less technical adopters aren’t catching up as fast. This might be a channel for skill-biased technological change — the people who benefit most from AI are those who already had advantages. Combined with the finding that global usage inequality is actually increasing (the Gini across countries rose from 0.46 to 0.50), AI might widens existing economic gaps rather than closing them.
And Anthropic opensourced the data in HuggingFace! Cheers.