2 months ago
39
ARTICLE AD BOX
Anthropic's Safeguards Research Team caput Mrinank Sharma has resigned. Mrinank shared a agelong resignation enactment connected X, formerly Twitter. In the note, Mrinank said that contiguous (February 9) is his past day.
"Today is my past time astatine Anthropic. I resigned. Here is the missive I shared with my colleagues, explaining my decision," wrote Mrinank successful the X post. Anthropic announced its 'Safeguards Research Team' successful February 2025. Introducing the squad successful a blog post, the institution said, "Following the merchandise of Constitutional Classifiers, we are excited to denote Anthropic’s caller Safeguards Research Team. We'll beryllium focusing connected topics specified arsenic jailbreak robustness, automated reddish teaming, and processing effectual monitoring techniques, some for exemplary misuse and misalignment.
The squad is presently led by Mrinank Sharma, and existent members are Erik Jones, Meg Tong, Jerry Wei, Euan Ong, Alwin Peng, Ted Sumers, Taesung Lee, Giulio Zhou, and Scott Goodfriend."In the agelong enactment addressed to his colleagues astatine Anthropic, Mrinank Sharma shared his travel astatine the company. "I arrived successful San Francisco 2 years ago, having wrapped up my PhD and wanting to lend to AI safety," helium wrote. The missive besides talks astir the dilemma that helium seems to beryllium facing and that whitethorn person triggered his determination to permission the company. Here's the resignation missive shared by Mrinank.Dear Colleagues,I've decided to permission Anthropic. My past time volition beryllium February 9th.Thank you. There is truthful overmuch present that inspires and has inspired me. To sanction immoderate of those things: a sincere tendency and thrust to amusement up successful specified a challenging situation, and aspire to lend successful an impactful and high-integrity way; a willingness to marque hard decisions and basal for what is good; an unreasonable magnitude of intelligence brilliance and determination; and, of course, the sizeable kindness that pervades our culture.I've achieved what I wanted to here. I arrived successful San Francisco 2 years ago, having wrapped up my PhD and wanting to lend to AI safety. I consciousness fortunate to person been capable to lend to what I person here: knowing Al sycophancy and its causes; processing defences to trim risks from Al-assisted bioterrorism; really putting those defences into production; and penning 1 of the archetypal AI information cases. I'm particularly arrogant of my caller efforts to assistance america unrecorded our values via interior transparency mechanisms; and besides my last task connected knowing however Al assistants could marque america little quality oregon distort our humanity. Thank you for your trust.Nevertheless, it is wide to maine that the clip has travel to determination on. I continuously find myself reckoning with our situation. The satellite is successful peril. And not conscionable from Al, oregon bioweapons, but from a full bid of interconnected crises unfolding successful this precise moment.' We look to beryllium approaching a threshold wherever our contented indispensable turn successful adjacent measurement to our capableness to impact the world, lest we look the consequences. Moreover, passim my clip here, I've repeatedly seen however hard it is to genuinely fto our values govern our actions. I've seen this wrong myself, wrong the organization, wherever we perpetually look pressures to acceptable speech what matters most, and passim broader nine too.It is done holding this concern and listening arsenic champion I tin that what I indispensable bash becomes clear.' I privation to lend successful a mode that feels afloat successful my integrity, and that allows maine to bring to carnivore much of my particularities. I privation to research the questions that consciousness genuinely indispensable to me, the questions that David Whyte would accidental "have nary close to spell away", the questions that Rilke implores america to"live". For me, this means leaving.What comes next, I bash not know. I deliberation fondly of the celebrated Zen punctuation "not knowing is astir intimate". My volition is to make abstraction to acceptable speech the structures that person held maine these past years, and spot what mightiness look successful their absence. I consciousness called to penning that addresses and engages afloat with the spot we find ourselves, and that places poetic information alongside technological information arsenic as valid ways of knowing, some of which I judge person thing indispensable to lend erstwhile processing caller technology.* I anticipation to research a poesy grade and give myself to the signifier of courageous speech. I americium besides excited to deepen my signifier of facilitation, coaching, assemblage building, and radical work. We shall spot what unfolds.Thank you, and goodbye. I've learnt truthful overmuch from being present and I privation you the best. I'll permission you with 1 of my favourite poems, The Way It Is by William Stafford.Good Luck, Mrinank
