Anthropic is an American AI start-up that specializes in the development of general AI systems and language models. It is also a public-benefit corporation with a mission to use artificial intelligence responsibly.
History of Anthropic
Anthropic was founded in 2021 by Dario and Daniela Amodei, a brother-and-sister team who are also former employees of OpenAI. The pair started the company with nine other OpenAI co-workers who shared a coherent vision for the future of AI research and safety.
In a podcast interview with the Future of Life Institute, Dario noted that “we [the team] were all working at OpenAI and trying to make this focused bet on basically scaling plus safety or safety with a lens towards scaling being a big part of the path to AGI… [we] just eventually came to the conclusion that it would be great to have an organization like top to bottom that was just focused on this bet and could make all strategic decisions with this bet in mind.”
This bet means Anthropic wants to make powerful AI systems less unpredictable, unreliable, and opaque. In more specific terms, the company’s objective is to incorporate safety principles into an AI industry that, like many others, favors power, efficiency and increasingly profit above all else.
Series A and Series B funding
Just months after the company was founded, it managed to raise $124 million in a Series A funding round led by Skype co-founder Jaan Tallinn. Also involved were Dustin Moskovitz, James McClave, former Google CEO Eric Schmidt, and the Center for Emerging Risk Research.
In a post on its website announcing the deal, Anthropic noted the funds would be used to research and develop AI systems that were interpretable, steerable, reliable, and robust.
This was followed by a $580 million Series B funding round in April 2022 that was led by FTX founder Sam Bankman-Fried. Daniel Amodei noted the money would be used to explore the scaling properties of machine learning systems and how safety issues might emerge at scale.
Google has subsequently invested nearly $400 million in Anthropic, and while both parties have declined to comment on the specifics, Anthropic will use Google’s cloud services to enable other businesses to harness the power of reliable and responsible AI.
Claude is the name of Anthropic’s language model assistant and the main rival to ChatGPT. Claude is a 52-billion parameter autoregressive model that was trained unsupervised on a large text dataset similar to its competitor.
While there are many similarities between Claude and ChatGPT, the former is trained on “Constitutional AI” principles and examples to make what Anthropic believes is a less harmful language assistant.
In February 2023, the company introduced a waitlist for early access to Claude.
In addition to Dario and Daniela Amodei, there are now over 50 employees at Anthropic. However, the company’s founding team is believed to consist of:
- Tom Brown – former lead engineer of GPT-3.
- Chris Olah – who worked at Google Brain on scaling laws.
- Sam McCandlish – former OpenAI research lead.
- Catherine Olsson – a machine learning researcher and computer scientist who is a technical staff member.
- Jack Clark – another Anthropic co-founder who worked in policy, strategy, and communications at OpenAI.
- Tom Henighan – like Olsson, Tom Henighan is a member of the technical staff who held a similar role at OpenAI.
- Kamal Ndousse, Benjamin Mann, and Nicholas Joseph – all three round out Anthropic’s technical staff team and like Henighan, held similar roles at OpenAI.
- Anthropic is an American AI start-up that specializes in the development of general AI systems and language models. It is also a public-benefit corporation with a mission to use artificial intelligence responsibly.
- Anthropic wants to make AI systems less unpredictable, unreliable, and opaque. To that end, the company’s objective is to incorporate safety principles into an AI industry that, like many others, prioritizes power and efficiency above all else.
- Claude is the name of Anthropic’s language model assistant and the main rival to ChatGPT. Claude is a 52-billion parameter autoregressive model that was trained unsupervised on a large text dataset similar to its competitor.
Connected Business Model Analyses
Stability AI Ecosystem