Anthropic: Redefining Artificial Intelligence with Ethics at its Core

Founded by former OpenAI executives and strongly backed by Amazon, Anthropic has developed an AI named Claude. It seeks to set itself apart from ChatGPT or Gemini in unexpected areas. Starting in 2015, many tech figures, including Elon Musk and Bill Gates, began voicing their concerns about artificial intelligence. As it pervades sectors like healthcare, military, or surveillance, any mistake could have severe consequences.

The Amodeis

In 2019, three years before the release of ChatGPT, these concerns affected two vice presidents of OpenAI: Dario and Daniela Amodei (siblings). As early as 2016, Dario Amodei voiced his concerns in an article published in collaboration with Google researchers titled “Concrete Problems in AI Safety”, worried about the unpredictability inherent in large-scale AI models. OpenAI was initially founded as a non-profit organization with the mission to “build a safe general AI and share the benefits with the world.” However, in 2019, it received a substantial investment of $1 billion from Microsoft. From then on, OpenAI became interested in the prospect of generating profits for its investors. The potential dangers of AI were no longer a primary concern for the leaders. This situation generated various internal tensions. As Dario Amodei testified: “We were a group within OpenAI who, after creating GPT-2 and GPT-3, firmly believed in certain ideas. The first was that if more computing power was devoted to these models, there would be virtually no limit. The second was that beyond improving the models, we had to consider their ethical alignment and safety issues.”

Founding of Anthropic

These issues related to the civilizational role of AI led Dario Amodei to leave OpenAI in December 2020. In the following weeks, 14 other researchers joined him, including his sister Daniela. The Anthropic foundation was launched soon after, in January 2021. Its ambition is stated as: “develop an extremely powerful but safe AI for the future of humanity.”

Constitutional AI

Anthropic’s endeavor to integrate ethical principles into AI and to mitigate unpredictable, unreliable, and opaque elements is summarized by a name: Constitutional AI. It incorporates 10 guiding principles ensuring that its responses are accurate, ethical, and beneficial to users, and must improve over time, respecting the various principles established by humans.

Early Funding

To develop an AI model, funds are needed. Anthropic rapidly attracted funding. The first – $124 million – was provided by Estonian billionaire Jaan Tallinn, co-founder of Skype. In October 2021, it was Sam Bankman-Fried’s turn, founder of the cryptocurrency platform FTX, to contribute $500 million to Anthropic – as we know, FTX would collapse a year later. Towards the end of the year, PitchBook, a platform that tracks private investment data, raised $704 million.

Birth of Claude

It wasn’t until the summer of 2022 that Anthropic unveiled the first version of its generative AI: Claude. The Amodeis chose not to release it, citing the necessity for further safety tests. Instead, OpenAI’s ChatGPT would reveal to the world the potentials of artificial intelligence. At the end of 2022, $300 million was invested by Google, securing a 10% stake in the startup. This investment facilitated access to significant computing resources. It wasn’t until March 2023 that Anthropic introduced the first version of Claude. To differentiate it from ChatGPT, the Amodeis claimed that their chatbot refused to generate harmful or biased content and provided sources to back its responses. In July 2023, a new version, Claude 2, surprised with its creativity and sense of humor. During that same month, Dario Amodei testified before the U.S. Senate to explain how unrestricted AI could aid in creating weapons of mass destruction. The major event took place in September when Amazon invested $1.25 billion in Anthropic, pledging to raise the amount to $4 billion. The work of the Amodeis was recognized by Time 100, naming them among the most influential people in the world in 2023.

Strong Valuation

The year 2024 began with a significant development. In February, venture capital firm Menlo Ventures agreed to invest $750 million in Anthropic, raising its valuation to $18.4 billion. Then, in June of the same year, Anthropic launched the Claude 3 family – Opus, Sonnet, and Haiku. This time, the model was acclaimed as one of the top on the market. In a new development, Anthropic and OpenAI agreed to submit their new models to the U.S. government in response to growing concerns about AI safety and ethics. Indeed, the National Institute of Standards and Technology (NIST) announced that it would now have “access to major new models from each company before and after their public release.” By emphasizing AI safety and ethics, Anthropic has initiated a significant movement in the history of this technology.