Google wants to help make Anthropic the next OpenAI

chatbot, chat, robot

OpenAI is the hot company. The launches of DALL-E 2 and ChatGPT have made it an absolute reference in an industry that promises to revolutionize our world. It is not alone, of course, and among its potential competitors is a company with big plans and especially strong support.

Anthropic. In 2021, several OpenAI veterans began to have differences with the strategy that the company was proposing. They ended up leaving it to create a new startup they called Anthropic. Among its founders are the brothers Daniela and Dario Amodei, the latter being former Vice President of Research at OpenAI.

Google is behind. The search giant maintains a much more cautious attitude than Microsoft when it comes to launching products based on generative AIs, but it continues to make moves. One of the last was precisely an investment of close to 400 million dollars in this company, thus becoming one of the main supports for the future. The move is in the same direction as Microsoft’s much larger investment in OpenAI. As part of the deal, Anthropic agreed to make Google its “preferred cloud provider.”

Claude. In January 2023 Anthropic launched ‘Claude’ , its particular rival to ChatGPT. Its behavior in certain scenarios was, according to some analysts, superior, although OpenAI would later release GPT-4. The name of this generative AI model, by the way, seems to be a tribute to Claude Shannon, ” the man who created the information “.  Access can be requested from Claude, though he is also available as the base of the Poe chatbot launched by Quora.

Anthropic wants to unseat OpenAI. Documents obtained by TechCrunch reveal the future plans of the company, which wants to raise a total investment of 5,000 million dollars in the next two years to rival OpenAI and thus be able to enter more than a dozen segments with the help of their developments in AI.

Claude-Next. This is the name of that next-generation model or “frontier model” (“frontier model”) that according to those documents will be 10 times more powerful than the most powerful AI currently. That will cost money, of course: at least $1 billion to be spent over the next 18 months. In this development, the so-called ” constitutional AI ” will play a prominent role, which according to those responsible for Anthropic helps to avoid risks and threats in the development of an AI that is aligned with human values.

They will need thousands of GPUs. In order to train such a “frontier model” Anthropic will need a computing power of 10^25 FLOPs, which is several orders of magnitude higher than what is used in major models today. To achieve this they will need “tens of thousands of GPUs”, the document indicates.

All to prove. The document marks a strategy and demonstrates Anthropic’s ambition, but this sector is in full swing. In addition to Google —which also has Google Brain and Deepmind— and Microsoft with OpenAI, giants like Meta, NVIDIA or Baidu are betting on AI.

Leave a Reply