Mrinank Sharma explains why he is leaving Anthropic and his new project
The researcher who led the team to protect AI systems at Anthropic has made public an internal communication. In it, he details the reasons for his departure and reveals what his next professional step will be in the field of algorithmic security. 🧠
Reasons behind the decision to leave
Sharma states that his departure is linked to a fundamental disagreement about the strategy to ensure that artificial intelligence is safe. He believes that the speed to develop these systems exceeds our ability to verify that they do not pose a danger. Therefore, he chose to focus his efforts from a more autonomous and independent position.
Key points of his argument:- There are differences in approach on how to prioritize safety in AI development.
- The current pace of technological advancement is faster than our capacity to implement safeguards effectively.
- He prefers to address the problem from an external framework with greater freedom of action.
"Sometimes, the safest way to protect a sandcastle is to step away from the shore and build a dike further away."
The future: a new non-profit organization
The former director announced that he is going to found an entity dedicated to analyzing the dangers posed by the most complex AI models. His goal is to evaluate and reduce potential large-scale threats, working with other experts to define more robust protocols.
Objectives of the new initiative:- Create an organization focused on researching risks of advanced AI.
- Collaborate with the research community to establish robust safety standards.
- Mitigate potential threats that may arise on a global scale.
A look ahead
The departure of a key figure in AI safety from Anthropic underscores the existing debates in the industry. Sharma's new project seeks to fill a gap, promoting independent evaluation of risks while technology continues its accelerated evolution. 🔍