Artificial intelligence research startup Anthropic revealed today that it has raised $450 million in new funding led by Spark Capital to drive its continued work towards building helpful and trustworthy AI systems, including its Claude AI assistant that can perform conversational and text processing tasks.

Launched in 2020, Anthropic has risen to become a well-known generative AI startup. Co-founded by the siblings Dario Amodei and Daniela Amodei who were vice presidents at the AI company OpenAI LP, it’s best known for the generative AI chatbot ChatGPT, which made the technology famous for its conversational capabilities.

This Series C funding round comes just two months after the previous $300 million round, also led by Spark Capital. The current round was joined by Google LLC, Salesforce Ventures and Zoom Ventures. The company did not disclose its valuation, but previously The Information reported that the company was aiming for $4.1 billion, so it may be close to that.

Anthropic Chief Executive Dario Amodei said these investments will support Anthropic’s primary mission of putting AI safety first. “The systems we are building are being designed to provide reliable AI services that can positively impact businesses and consumers now and, in the future,” he said.

Zoom partnered with Anthropic earlier this month to integrate Anthropic’s Claude AI assistant into its products, including the company’s customer-facing communications systems such as its contact center and support systems. With the partnership, customers can get their own self-service answers to questions from the AI assistant or employees can use the chatbot in order to assist them when on calls.

Competition between AI models in the market has been increasing over the months as new capabilities and research has exploded since ChatGPT was unveiled in November. Anthropic competes with numerous AI models in generative AI and text processing such as OpenAI ChatGPT, Microsoft Corp.’s Bing Chat — powered by OpenAI’s GPT-4 — Google LLC’s Bard and Cohere Inc.’s AI model. Other AI models exist that can generate artwork from text prompts such as DALL-E 2 and Stable Diffusion which have also emerged.

To stay ahead of the pack, Anthropic has focused on building safer models that are less likely to suffer from “hallucinations,” such as when a generative AI goes off the rails and offers completely false information with complete confidence. The company has also worked on implementing guardrails against toxic behavior such as sexism and racism in its models to further safety research in AI handling. To do this, Anthropic trains its AIs to choose the least harmful replies to make them the most “helpful, polite, respectful and thoughtful,” in order to avoid dangerous behaviors.

“Our team is focused on AI alignment techniques that allow AI systems to better handle adversarial conversations, follow precise instructions, and generally be more transparent about their behaviors and limitations,” Anthropic wrote in the announcement.

Part of this research included increasing the amount of data that Claude, and Anthropic’s other AI models, can digest at once up to 100,000 tokens. That allows the AI to hold longer conversations without getting confused or lost – and it also means that it can read much longer documents, such as books or business proposals to summarize them or provide insights. To test the new system, the team had it read “The Great Gatsby.”

“Anthropic has assembled a world-class technical team that is dedicated to building safe and capable AI systems,” said Yasmin Razavi, general partner at Spark Capital. “The overwhelmingly positive response to Anthropic’s products and research hints at AI’s broader potential for unlocking a new paradigm.”

Image: Anthropic