Alex Tamkin

Alex Tamkin: Which Economic Tasks are Performed with AI? On May 19, 2025, Alex Tamkin, research scientist at Anthropic, will stop by the lab for our seminar series.

This Monday 5/19, @alextamkin.bsky.social of @anthropic.com will stop by the Lab for our seminar series!

Details and registration here: digitaleconomy.stanford.edu/event/alex-t...

@caseynewton.bsky.social on the latest @anthropic.com Economic Index reporter authored by @alextamkin.bsky.social: www.platformer.news/people-are-u...

Save the date: @alextamkin.bsky.social joins @beckerfriedman.bsky.social economists and Chicago’s former Deputy Mayor for Economic Development Samir Mayekar on 4/15 to discuss AI’s impact on the economy and the @anthropic.com Economic Index.

"gets rave reviews from clients"

I'm here in DC today to talk about Anthropic's Economic Index research on AI and work on a panel moderated by @ashleyrgold.bsky.social at AEI.

Come join in-person or tune in virtually: www.aei.org/events/ai-an...

@alextamkin.bsky.social in @washingtonpost.com today re: Claude usage we're seeing in occupation-linked tasks from our @anthropic.com Economic Index research: www.washingtonpost.com/business/202...

Thanks to everyone else who gave feedback and helped out along the way.

There's lots more to come, so stay tuned.

6/6🌅

This work was a huge group effort.

Thanks to my amazing coauthors, including co-lead Kunal Handa, Miles McCain, Saffron Huang, Esin Durmus, Sarah Heck, Jared Mueller, Jerry Hong, Stuart Ritchie, Tim Belonax, Kevin K. Troy, Dario Amodei, Jared Kaplan, Jack Clark, and Deep Ganguli

5/

Anthropic's Economic Data: Feedback & Interest Form We have developed a privacy-preserving dataset from millions of human-AI interactions across economic tasks. Our aggregated and anonymized dataset provides unique insights into how AI systems are bein...

Third, there are many limitations to our work! (See Section 4.1 in the paper). We're working on making progress on them.

We’re also excited to see what others do with our data—please get in touch via this form with any feedback or input:

docs.google.com/forms/d/e/1F...

4/

Second, because our methods are fully automated, we can run the same process over time on a recurring basis.

This gives us a moving picture of how AI is advancing across the economy, and identifies leading indicators that we can use to plan.

3/

A horizontal bar chart comparing Augmentation vs Automation of tasks, showing percentage of conversations. The Augmentation bar (57% total) is split into three categories: Validation (2.8%), Task iteration (31.3%), and Learning (23.3%). The Automation bar (43% total) is divided into two categories: Feedback loop (14.8%) and Directive (27.8%). The bars use different shades of blue for Augmentation and purple for Automation categories. The graph suggests AI is used slightly more for augmenting human tasks than for automation.

While those links have all the details, I wanted to call out two additional points:

First, I’m excited that our methods give insight into *how* work is done with AI

There’s a lot of interest in *what* tasks AI will do, but *how* AI changes the nature of work is also crucial.

2/

A title card with dark text on a cream background reading 'Which Economic Tasks are Performed with AI? Evidence from Millions of Claude Conversations' by Handa & Tamkin et al. The Anthropic logo appears in the bottom left. On the right is a black and white macro photograph of a worker bee on a honeycomb.

How is AI being used across the economy?

We have some new research and datasets to share:

Paper: assets.anthropic.com/m/2e23255f1e...

Data: huggingface.co/datasets/Ant...

Blogpost: anthropic.com/news/the-ant...

1/

Title card: Alignment Faking in Large Language Models by Greenblatt et al.

New work from my team at Anthropic in collaboration with Redwood Research. I think this is plausibly the most important AGI safety result of the year. Cross-posting the thread below:

How are AI Assistants being used in the real world?

Our new research shows how to answer this question in a privacy preserving way, automatically identifying trends in Claude usage across the world.

1/

In what the company is calling a first for a major AI lab, the Clio paper also highlights the top three categories of uses for Claude: Coding and software development (more than 10 percent of conversations) Educational use, both for teachers and for students (more than 7 percent) Business strategy and operations, such as drafting professional communications and analyzing business data (almost 6 percent)

Clio is Anthropic's new system for identifying AI risks that it hadn't thought to look for — what it calls the unknown unknowns. I talked with team that built it and share for the first time the top three ways people use Claude www.platformer.news/how-claude-u...

How Claude uses AI to identify new threats PLUS: Exclusive data on how people are using Anthropic’s chatbot

For more, check out this article from @caseynewton.bsky.social on our work and findings!

www.platformer.news/how-claude-u...

5/5

Clio: Privacy-preserving insights into real-world AI use A blog post describing Anthropic’s new system, Clio, for analyzing how people use AI while maintaining their privacy

We also talk at length in the blogpost and paper about how Clio works, its privacy measures, and how its insights can help us improve our current and future safety systems:

www.anthropic.com/research/clio

Paper: assets.anthropic.com/m/7e1ab885d1...

4/

And some insights into how Claude use varies across different languages

3/

For example, here are the most common use cases on Claude.ai…

2/

How are AI Assistants being used in the real world?

Our new research shows how to answer this question in a privacy preserving way, automatically identifying trends in Claude usage across the world.

1/

Congrats!!

Eventually your learning rate slows!

Aerial picture of the UBC campus, with an arrow pointing to a building and text asking "Your PhD lab?"

Do you want to understand how language models work, and how they can change language science? I'm recruiting PhD students at UBC Linguistics! The research will be fun, and Vancouver is lovely. So much cool NLP happening at UBC across both Ling and CS! linguistics.ubc.ca/graduate/adm...

Some really cool interpretability work on protein language models!

www.biorxiv.org/content/10.1...