- Anthropic, founded two years ago by former OpenAI research executives, unveiled its new AI chatbot called Claude 2 and invited the public to use it.
- The company has raised $750 million in two rounds of funding since March and is valued at $4.1 billion.
- In May, Anthropic was one of four companies invited to a White House meeting to discuss responsible AI development with Vice President Kamala Harris.
Anthropic’s Dario Amodei, right, arrives at the White House for a meeting with Vice President Kamala Harris on artificial intelligence on Thursday, May 4, 2023 in Washington.
Evan Vucci | AP
There’s a new addition to the burgeoning AI arms race.
As Microsoft-backed OpenAI and Google compete to develop the most advanced chatbots powered by generative artificial intelligence, Anthropic is investing heavily to keep up. Just months after raising $750 million in two rounds of funding, the startup is introducing a new AI chatbot: Claude 2.
Anthropic was founded in 2021 by former OpenAI research executives and funded by the likes of Google, Salesforce, and Zoom. With Claude 2, Anthropic is making its chatbot technology accessible to consumers for the first time. Over the past two months, the company’s AI models have been tested by companies including Slack, Notion and Quora, and Anthropic has built a waiting list of more than 350,000 people requesting access to Claude’s API and consumer offering.
“We were focused on business and on making Claude as safe as possible,” said Daniela Amodei, who co-founded Anthropic with her brother Dario. “We genuinely believe this is the most secure version of Claude we’ve developed to date, so we’re excited to make it available to a wider range of businesses and consumers.”
Claude 2 will initially only be available to US and UK users, and Anthropic plans to expand availability in the coming months.
Since OpenAI unveiled ChatGPT to the public late last year, the tech world has invested heavily in the potential of generative AI chatbots that respond to text prompts with sophisticated and conversational responses. Academics and ethicists have raised significant concerns about technology’s tendency to spread bias, yet it has quickly taken hold in schools, online travel, the medical industry, online advertising, and more.
In March, OpenAI released GPT-4, its biggest update to the underlying technology behind ChatGPT. Two months later, ChatGPT allowed web browsing to begin, so responses were no longer limited by the 2021 training data cut-off date. However, the chatbot’s web browsing ability was subsequently disabled following reports of problematic uses such as bypassing paywalls.
Google announced its Bard competitor in February and has since updated the chatbot’s math and programming capabilities and integrated the AI service into more products.
Anthropic’s ambitions are no less lofty.
The company says Claude 2 can summarize up to 75,000 words, which could be the length of a book. Users can enter large data sets and request summaries in the form of a memo, letter, or story. ChatGPT, on the other hand, can handle around 3,000 words.
Picture Alliance | Picture Alliance | Getty Images
Daniela Amodei said Anthropic invested at least two months in the development of Claude 2, with a team of 30 to 35 people working directly on the AI model and a total of 150 people supporting it. She said the market is growing so fast that there is plenty of room for multiple players to succeed.
“It’s a really unusual time from a business perspective because the demand for large language models is just so great and actually more demand than the industry can currently provide,” Amodei said. “The landscape is just very broad and there’s really a lot of room for a lot of different users and user types to use these systems.”
In May, Anthropic was one of four companies invited to a White House meeting to discuss responsible AI development with Vice President Kamala Harris. The others were Google parent company Alphabet, Microsoft and OpenAI.
That same month, the company raised $450 million after raising a $300 million round of funding in March at a valuation of $4.1 billion.
Amodei provided some concrete examples of the Claude 2’s improvements over the previous version of the model. The new chatbot scored 76.5% on the multiple choice section of the bar exam, up from 73%. And in a Python coding test, Claude 2 scored 71%, up from 56% in the previous iteration.
When it comes to security, the updated chatbot was twice as good at providing “harmless answers,” according to a blog post, although it’s worth noting that the red team’s security assessment was conducted in-house.
Despite Claude 2’s performance improvements, Amodei acknowledged that Anthropic and the wider industry still have many hurdles to overcome. For example, the tendency of AI chatbots to make up wrong answers, referred to as “hallucinations” by some tech companies, is an ongoing problem.
“Today with all language models in the world, there are just so many unknown unknowns but also known challenges,” Amodei said. “No language model is 100% immune to hallucinations, and Claude 2 is the same.”
REGARD: ChatGPT has boosted optimism about AI