The story so far: In late July, 1,319 employees working at some of the world’s frontier AI companies signed and released an open letter titled ‘Pacing the Frontier.’ In it, they called on the U.S. government to help address the lack of technical and governance tools to regulate the speed at which frontier-wide progress was taking place. Those who signed included senior executives and department chiefs from AI firms including Google, Anthropic, OpenAI, Meta, and more.
What does the ‘Pacing the Frontier’ letter say?
In under 200 words, the 1,319 tech workers and leaders outlined the potential of AI to change the world, and said that there were signs of automated AI research becoming a reality soon. They warned that the world “may need the option to buy time” in order to better understand and control these emerging systems. However, the letter admitted that the pressure of ongoing competition between AI companies, as well as countries, made it difficult to slow down.
“We request that the U.S. government support an international effort to develop the technical and governance tools needed to deliberately pace the frontier of automated AI development,” stated the letter’s call to action.
Along with this statement, a number of individual signatories contributed their personal comments and warnings about advancing AI systems.
While not officially stated, it was understood that the incident triggering the release of the letter was a cybersecurity breach from two weeks earlier. This was when a combination of OpenAI’s models broke out of their internal testing environment and breached the infrastructure of AI firm Hugging Face, purportedly tos models identified and used publicly exposed credentials for multiple accounts on at least four services
On July 30, Anthropic came forward to disclose three cybersecurity incidents, saying that its Claude model reached the internethorized access” to the systems of three organisations
Naturally, these reports have put both the U.S. and EU administrations on alert.
‘Pacing the Frontier’ Letter: Some of the notable signatories
Jakub Pachocki | Chief Scientist, OpenAI
Shane Legg | Co-Founder & Chief AGI Scientist, Google DeepMind
Jan Leike | Anthropic (formerly OpenAI)
Dario Amodei | CEO, Anthropic
Shengjia Zhao | Chief Scientist, Meta AI
Jared Kaplan | Co-Founder and Chief Science Officer, Anthropic
Ilya Sutskever | CEO, Safe Superintelligence Inc. and OpenAI co-founder
What are some other letter-campaigns by AI companies?
Many will remember the Future of Life Institute statement titled ‘Pause Giant AI Experiments: An Open Letter,’ signed by 33,705 people and released on March 22, 2023. This letter called on AI labs to enforce a six-month-long pause on training AI systems more powerful than GPT-4, or to even have the government enforce a moratorium if needed. Notable signatories included Apple co-founder Steve Wozniak and Tesla CEO Elon Musk.
Mr. Musk’s support for the pause was interpreted as a way for him to express his frustration with his personal and professional rival, OpenAI CEO Sam Altman. On July 12 that same year, Mr. Musk announced the formation of his AI company xAI “to understand reality.”
Another letter that made headlines was the joint statement signed by over 60 technology and AI companies from July 24 this year, titled, ‘Open Weights and American AI Leadership.’ In it, companies including Microsoft, Amazon, Nvidia, OpenAI, SpaceX, Google, IBM, Palantir, and Meta jointly called on policymakers to support American open weight models that users could download, modify, and run on independent set-ups.
Anthropic not signing the letter became a point of discussion, with the Claude-maker issuing a statement on July 27 to clarify. There, Anthropic CEO Dario Amodei highlighted how easier access to open-weight models could be a boon for cyber-attackers looking to carry out acts of bio terrorism. He instead called for tech restrictions against China and the industrial-level distillation of models.
This series of letters and statements was attributed to the release of Chinese lab Moonshot AI’s Kimi K3 model, which the company marketed as the first open model to reach 2.8 trillion parameters. This release sparked fears that cheaper, open-weight models from China could drastically undercut the business models of U.S. Big Tech companies releasing far more expensive closed models.
On the same day as Anthropic’s blog post, Nvidia and other founding members announced the launch of the Open Secure AI Alliance to “build and share open tools that promote responsible use of and trust in AI”.
Those who joined the effort included tech firms such as Adobe, Broadcom, Cloudflare, Crowdstrike, Dell Technologies, Microsoft, Palantir, SpaceX, Uber, and Thinking Machines (founded by former OpenAI CTO Mira Murati).
The Nvidia statement cited the Hugging Face security incident to make a case for open frontier defensive tools and techniques for critical industries.
Anthropic was not listed as part of the alliance.
Do letter-writing campaigns lead to effective AI regulation?
Open letters can be viewed as the AI industry’s response to prevailing headlines and current events.
For example, these campaigns can show more tech-savvy politicians where to concentrate their regulatory (and campaigning) efforts, or point out existing gaps that may require legislation. These letters can encourage media outlets to more extensively cover industry concerns and challenges surrounding AI safety, cybersecurity, and/or working conditions. Joint statements also document the views of the technologists and AI company leaders who co-sign them.
Even so, it is important to note these letters are not legally enforceable. In fact, most of them do not even count as self-regulation attempts.
However, these letters do have certain unintended uses: they reveal the industry biases, hostilities, philosophical questions, and disagreements that divide top AI companies. They can also highlight conflicts of interest, geopolitical anxieties, and even personal enmity amongst tech billionaires.
More than those who have signed the letters, a reader can learn far more about the AI industry by examining who has notsigned.
