Anthropic, an AI research startup, plans to secure up to $5 billion in the next two years to compete with rival OpenAI and enter over a dozen key industries, according to Techcrunch Report.
These and other long-term objectives for Anthropic, which was founded in 2020 by former OpenAI researchers, are revealed in a pitch deck for the company’s Series C fundraising round.
Anthropic says in the deck that it is developing a “frontier model” tentatively dubbed “Claude-Next” that is 10 times more capable than today’s most powerful AI, but that this will cost a billion dollars over the next 18 months.
When contacted for comment, an Anthropic representative stated that the company is planning additional product announcements and will be discussing them soon.
In early March, The Information stated that Anthropic was looking to secure $300 million at a valuation of $4.1 billion, bringing its total value to $1.3 billion.
The deck affirms that target number, though only half was made from a “confidential investor” at the time the document was created.
Anthropic describes the frontier model as a “next-generation algorithm for AI self-teaching,” referring to an AI training technique called “constitutional AI” that it developed. At a high level, constitutional AI seeks to align AI with human intentions by allowing systems to answer inquiries and perform tasks based on a simple set of values that guide them.
Anthropic estimates that its frontier model will need 1025 FLOPs, or floating point operations several orders of magnitude larger than even the most powerful models currently available.
Moreover, how this translates to computation time is dependent on the speed and scale of the system performing the computation; Anthropic suggests (in the deck) that it depends heavily on clusters with “tens of thousands of GPUs.”
This frontier model could be used to create virtual assistants that can reply to emails, perform research, and generate art, books, and other content, similar to what we’ve seen with GPT-4 and other large language models.
According to the pitch deck, “these models could begin to automate large portions of the economy.” “We believe that companies that train the best 2025/26 models will be too far ahead in subsequent cycles for anyone to catch up.”
The frontier model is the successor to Claude, Anthropic’s chatbot that can be programmed to perform a variety of tasks such as document searching, summarizing, writing and coding, and responding to questions about specific topics.
It’s similar to OpenAI’s ChatGPT in these ways. However, Anthropic argues that thanks to constitutional AI, Claude is less likely to produce harmful outputs, is simpler to engage in conversation with, and is more steerable.
Following a closed beta testing close to the end of last year, Anthropic launched Claude commercially in March, with around 15 partners receiving early access.
Former OpenAI VP of Research Dario Amodei founded Anthropic as a public benefit corporation in 2021, bringing with him a number of OpenAI employees, including OpenAI’s former policy lead Jack Clark. Amodei left OpenAI due to a disagreement about the company’s direction, specifically the startup’s growing commercial concentration.
Anthropic presently competes with OpenAI and startups such as Cohere and AI21 Labs, all of which are developing and commercializing their own text-generation and, in some cases, image-generation AI systems.
The pitch deck states that the company has created an approach for go-to-market and initial product specialization that fits with its core competency, brand, and where it sees adoption happening over the next 12 months.
According to the pitch deck, Alameda Research Ventures, the sister firm of Sam Bankman-defunct Fried’s cryptocurrency startup FTX, was a “silent investor” in Anthropic, with “non-voting” shares responsible for driving the $580 million Series B round. Anthropic anticipates that Alameda’s shares will be liquidated in bankruptcy proceedings within the next few years.
Google has also promised $300 million in the copany in exchange for a 10% stake in the startup. Anthropic agreed to make Google Cloud its preferred cloud offer under the terms of the agreement, which was first reported by the Financial Times, with the companies co-develop[ing] AI computing systems.
James McClave, Facebook and Asana co-founder Dustin Moskovitz, former Google CEO Eric Schmidt, and Skype co-founder Jaan Tallinn are among the other Anthropic backers.
Don’t miss important articles during the week. Subscribe to techbuild.africa weekly digest for updates.