Bergur Thormundsson
Research expert covering technology in the Nordics, AI and smart cities
Anthropic - statistics & facts
Anthropic is one of the leading developers of large language model artificial intelligence (AI), more commonly termed generative AI. Their model, Claude, competes with the most advanced AI models. The startup is relatively young, founded by a group of AI developers that left OpenAI in 2021. This new enterprise has been successful at attracting capable staff and in May 2024, more former employees of OpenAI joined the company, citing safety concerns at their old firm. Among its chief investors are Amazon and Google, with the latter also developing its own generative AI, Gemini.
Claude 3 with outstanding performance
Claude is on its third iteration, Claude-3, which was released in March 2024. Claude-3 itself is then broken down into three modules: Opus, the most advanced heavyweight of the modules, Sonnet, and Haiku. Sonnet and Haiku are medium and small-sized models, intended for lighter tasks. All three of the modules can also accept image input, meaning they form an early-stage multimodal generative AI model, one that can take in multiple types of datasets. Opus, however, is a rival to the two other most advanced generative AI programs, ChatGPT from OpenAI and Google’s Gemini. Though Anthropic claims Opus outperforms ChatGPT-4 and Gemini Ultra, when benchmarked, all three models are of very comparable capability. In terms of adoption Claude Opus has already been integrated into Amazon’s Bedrock, its web-based cloud AI services.AI is expensive and funding limited
Anthropic has done well in acquiring financial support and funding from a variety of sources. The most notable of these are Amazon and Google. Amazon invested a sizeable amount of nearly four billion dollars into the enterprise and Google has invested some 500 million dollars as well as committing to another 1.5 billion dollars going forward. This, however, might not be enough as the costs of running an AI model are immense and funding is not infinite. The profitability of generative AI models has also not been streamlined, with many burning up to a billion U.S. dollars per year while bringing in revenue measured only in a few hundred million.Tough times against tough competition
Anthropic’s open goal is to research the safety and reliability of AI. To this end, it has incorporated such documents as the 1948 Universal Declaration of Human Rights into Claude’s creation. However, this humanitarian angle may not shield them from the financial realities of running a costly AI system. And as Google operates Gemini, one of Claude’s toughest rivals, it is possible that funding might turn scarce in the coming years. Fixing the profitability of its generative AI models will be a critical step to allow Anthropic to continue operations for years to come.
