Confrontation with the Pentagon intensified competition among leading artificial intelligence developers

Contracts with the Pentagon have become a new point of tension in the artificial intelligence industry. The differing attitudes of two leading laboratories — OpenAI and Anthropic — towards cooperation with the U.S. Department of Defense led to a conflict that also affected the struggle for talent in the AI field.

OpenAI agreed to deploy its models in classified Pentagon systems, which caused a wave of dissatisfaction within the company itself.

As the WSJ writes, at least two top managers resigned from OpenAI after this deal: Vice President of Research Max Schwarzer and Head of Hardware Caitlyn Kalinowski. The latter openly stated that issues of citizen surveillance and the use of autonomous weapons require much deeper discussion than what was conducted before signing the contract.

At the same time, Anthropic took an unyielding stance. The company tried to include strict safeguards in its agreement with the Pentagon that would prohibit the use of its technologies for domestic espionage and the creation of lethal weapons. When negotiations reached an impasse, the Trump administration declared Anthropic a "supply chain risk".

Although this cost the company the loss of a major government client, in return it gained significant reputational benefit. This was confirmed not only by the fact that the Claude chatbot for the first time surpassed ChatGPT in App Store downloads, but also by unprecedented support from the public and industry colleagues. Thus, 37 employees of OpenAI and Google DeepMind even filed an appeal in court in support of Anthropic, which is now challenging the government's decision.

The financial aspect of this struggle looks no less impressive. As reported by the WSJ, last summer Meta corporation tried to lure away top engineers from OpenAI and Anthropic, offering compensation packages worth hundreds of millions of dollars.

In response to this challenge, OpenAI began paying bonuses and changed its stock allocation system to retain people. However, Anthropic chose a different path: the company's head Dario Amodei told employees that the payment structure would remain unchanged, as dedication to the mission is paramount. The result was telling: while several dozen people left OpenAI for Meta, Anthropic lost only two.

According to the author, Anthropic's vitality is explained by its founding history. The company was founded in 2021 by OpenAI alumni who believed the company was moving too fast with commercial products, paying insufficient attention to safety. Today, Anthropic prides itself on retaining almost all of its co-founders and most of its first 20 employees.

This success in talent retention forced OpenAI to go into damage control mode. Sam Altman publicly admitted that the hasty deal with the Pentagon looked "sloppy", and tried to appease the team through Q&A sessions, but, as seen by the departure of key researchers, these measures do not always work.

At the same time, as the WSJ writes, Anthropic also faces internal challenges. Security researcher Mrinank Sharma recently left the company. In a letter to colleagues, he wrote that he wanted to pursue poetry and expressed concern that the development of artificial intelligence could become a threat to the world. As the WSJ reported, his dismissal was partly related to changes in the company's security policy.

Thus, the story with the Pentagon contracts showed that in the artificial intelligence industry, not only money and technology play an important role. For many leading researchers, it is equally important how and for what purposes the systems they create will be used.

• 9
• 0
• 2
• 0
• 0
• 0
• 0
• 0