Anthropic, a San Francisco-based artificial intelligence startup founded by siblings Daniela and Dario Amodei, both former OpenAI executives, has raised $700 million in just a matter of weeks from tech giants like Google and Salesforce as investors continue to place large bets on generative A.I. technologies.
The two-year-old company just closed a $300 million funding round that valued it at $4.1 billion, the Information first reported yesterday (March 8). The round was led by Spark Capital, a venture capital firm with early stakes in Twitter and Tumblr. Salesforce is also a new investor, Anthropic disclosed in a tweet yesterday. Only four weeks ago, the company received an investment between $300 million and $400 million from Google, according to the Financial Times.
Anthropic’s main product is Claude, a ChatGPT-like A.I. chatbot released in January. Currently it’s only available to a group of test users.
Dario Amodei, the CEO of Anthropic, worked at OpenAI between 2016 and 2020 and led the development of the firm’s GPT-2 and GPT-3 language models, according to his LinkedIn profile. GPT-3 is the language model behind ChatGPT, which was released in November. Before OpenAI, Dario was a research scientist at Google and the U.S. branch of Chinese internet giant Baidu.
Daniela Amodei had a shorter career at OpenAI where she led the company’s recruiting effort and managed its technical safety and policy teams, according to her LinkedIn page. Before that, she worked in similar functions at payment software maker Stripe for five years. Daniela serves as the Anthropic’s president, overseeing the company’s day-to-day operations.
The Amodei siblings and other Anthropic cofounders chose to start their own company to focus on building what they call “steerable, interpretable and reliable A.I. systems with humans at the center of it,” Daniela said in an interview with the Future of Life Institute, a science nonprofit, in March 2022.
As (artificially) intelligent as they may seem, A.I. chatbots including ChatGPT have been criticized for producing factually wrong and biased answers to some prompts. Anthropic’s goal is to fix those shortcomings. “We are training large neural generative models and doing safety research on those models. We want to make those models safer and more aligned with human values,” Daniela Amodei said in the podcast.
In a white paper published in December, Anthropic said its work focuses on experimenting with methods for training a chatbot that learn to identify and combat harmful queries through self-improvement, rather than human intervention. “The only human oversight is provided through a list of rules or principles,” the paper said.
Before this year, Anthropic had raised another $700 million through two funding rounds in 2022, according to Crunchbase, a database of startup funding activities. Anthropic’s early investors include Facebook cofounder Dustin Moskovitz, former Google CEO Eric Schmidt and Sam Bankman-Fried, the disgraced founder of FTX Group.