The Anthropic Economic Index, brain-to-text, and a rocky market
Anthropic opens up data on AI's impact on labour, Mira Murati launches a lab, Meta decodes unspoken sentences, and Palantir tumbles. A week at the economics–AI intersection.
As someone interested in the intersection of economics and emerging technology, I was curious about Anthropic's announcement this week: the Anthropic Economic Index, a new initiative to track and analyse AI's impact on labour markets and the broader economy, derived from millions of anonymised Claude conversations.
The Anthropic Economic Index aims to understand AI's effects on the labour market and broader economy over time. It provides the clearest picture yet of how AI is being incorporated into real-world tasks across the modern economy.
Anthropic is open-sourcing the dataset to invite further research and policy input. What's genuinely cool is how it's addressing a wider, more interesting set of use cases beyond code co-piloting and technical writing — I'll be digging into the data.
Murati's new lab, and chip nerves
Mira Murati, former CTO at OpenAI, launched Thinking Machines Lab this week, assembling researchers and engineers from leading labs (most notably OpenAI) with a focus on human-AI collaboration. They aim to make AI 'more widely understood, customizable, and generally capable' than what's available today. In chip-land, as Nvidia's stock took a further hit, Jensen Huang addressed concerns about DeepSeek's R1: while some investors worried R1's efficiency might lower hardware demand, Huang reiterated that the model's compute-intensive nature would likely increase the need for advanced computing resources.
Enterprise governance ramps up
Singapore pledged S$150m to boost enterprise AI capabilities. And there's a clear step-up in GRC (Governance, Risk and Compliance) funding — Singulr raised a $10M seed for enterprise AI governance and Securiti announced a governance partnership with Databricks. I expect this space to only ramp up as organisations get over their year-of-fear and start genuinely operationalising long-term AI.
Brain-to-text, and a Palantir wobble
BrainBook? Meta has developed a non-invasive method to decode unspoken sentences from brain signals using MEG scanners. Musk is no longer the only broligarch developing neurotech. Terrifying or exciting, you choose; I'm coining it here: brain-to-text (BTT). And last but not least, Palantir shares tumbled mid-week, sparked by founder-CEO Alex Karp selling a chunk of his stock, likely tied to announced Pentagon budget cuts over the next five years. With a wider tech sell-off amid macro volatility, I'm holding for now — rocky, but not terminal.