Overview of the Internal Conflicts at OpenAI and the Formation of Anthropic

<p class="dropcap">In San Francisco during the mid-2020s, the atmosphere is charged with anticipation and uncertainty as engineers work on significant AI developments. The outcomes of these projects are believed to have major implications for global affairs and financial markets.</p><p dir="ltr">Dario Amodei, a physicist educated at Princeton, transitioned from Google to OpenAI in 2016. Known for his intense focus on AI safety, he authored a notable essay titled “Big Blob of Compute” in 2017, outlining his theory that machine intelligence is primarily a function of scale, influencing OpenAI's research direction towards large language models and the pursuit of artificial general intelligence (AGI).</p><p dir="ltr">By mid-2020, three main teams were competing towards AGI: OpenAI, DeepMind, and Google Brain, with Amodei leading a fourth team at OpenAI. Despite the success of GPT-3, Amodei grew disillusioned with OpenAI's leadership, particularly regarding their shift to a commercial model, which he felt deviated from the organization's original mission to benefit humanity.</p><p dir="ltr">Amodei expressed concerns about the leadership of Sam Altman and Greg Brockman, believing that their decisions were detrimental to the organization's ethical standards. He became particularly alarmed by a fundraising strategy known as the “countries plan,” which proposed selling AGI technology to governments, including the United States, China, and Russia. Amodei argued that this approach contradicted OpenAI's mission and could exacerbate global tensions.</p><p dir="ltr">In December 2017, Amodei communicated his objections to the countries plan in an email to OpenAI executives, emphasizing the potential risks associated with selling AGI to authoritarian regimes. Although the plan was eventually abandoned, Amodei's concerns about the organization's direction persisted.</p><p dir="ltr">Throughout 2018 and 2019, Amodei continued to voice objections to various business decisions at OpenAI, including the commercialization of GPT-3. His growing mistrust of Altman led him to document instances where he felt misled. He also raised ethical concerns regarding Altman's interactions with controversial figures.</p><p dir="ltr">In 2020, Amodei proposed to his colleagues the idea of forming a new company focused on AI safety, leading to the establishment of Anthropic. He and several co-founders left OpenAI to create a research organization that prioritized safety in AI development.</p><p dir="ltr">After negotiating their departure with OpenAI, Amodei and his team announced their new venture in a virtual meeting, where they were subsequently removed from the call and cut off from company resources.</p><p dir="ltr">Five years later, Anthropic has grown significantly, with thousands of employees and a focus on AI safety. The company has evolved from its initial vision into a major player in the AI field, competing closely with OpenAI.</p><p dir="ltr">Despite their growth, the Amodeis maintain a contentious relationship with OpenAI's leadership, characterized by mutual distrust and criticism. Both companies continue to navigate the complexities of AI development while grappling with ethical considerations and their competitive dynamics.</p><p dir="ltr"><em><small>This article is adapted from Roose’s book, </small></em><small><a href="https://bookshop.org/p/books/the-agi-chronicles-the-inside-story-of-the-race-to-create-an-artificial-superintelligence-kevin-roose/fedb4aef52bfc0d9?ean=9780374618759&amp;bkshp-astro=t">The AGI Chronicles: The Inside Story of the Race to Create an Artificial Superintelligence.</a></small></p>