"Work is only meaningful when it comes from love."
These words appear on Mrinank Sharma’s personal website.
He is a researcher with a PhD in statistical machine learning from Oxford. He is also a published poet, a meditation practitioner who leads a community in Berkeley, and a DJ who hosts dance gatherings to explore wisdom.
Two years ago, he joined Anthropic with a sense of mission: to protect humanity from the risks of AI. Leading the Safeguards Research Team, he studied sycophancy—AI’s tendency to flatter humans—and put a defense system against AI-enabled bioterrorism into production.
On February 9, he made his resignation letter public.
Dear colleagues,
I am leaving Anthropic.
I have continually confronted our situation. The world is in peril. This is not just about AI. Right now, interwoven crises are unfolding all at once.
The pressure to set aside what matters most is relentless.
I don’t know what comes next. A well-known Zen saying comes to mind.
"Not knowing is most intimate."
(Not knowing is most intimate)
Thank you. Goodbye.
Mrinank
Someone who worked on the front lines of AI safety is leaving. His parting words speak of how hard it is to put values into action.
Someone who builds AI studied how "AI could make us less human," and in the end decided to return to being a poet.
What was the thread he held on to? And what thread should we hold on to?
