The AI Arms Race Heats Up: Amazon Partners with Anthropic

Amazon and Anthropic this week announced a $4 billion partnership. What does this mean and what might the consequences be?

The AI ecosystem is developing with blistering speed. OpenAI, with its GPT-3 and GPT-4 large language models (LLMs) and its associated ChatGPT chatbot application, is perhaps the best-known AI company. However, four major tech companies -- Google/Alphabet/DeepMind, Facebook/Meta, Amazon and Microsoft -- are all deeply involved in what can be called an AI arms race.

The AI Arms Race Playing Field
Briefly, the current AI revolution is based on LLMs. An LLM is a relatively simple but huge neural-based software system that has been trained on a vast amount of text data, with sources typically including data troves like the entire contents of Wikipedia. The size of an LLM is measured by the number of parameters it has, 175 billion in the case of GPT-3. The more parameters an LLM has, the more powerful (human-like) it is. LLMs require huge computing resources and associated power consumption that can only be supplied by very large companies or government entities.

OpenAI and Microsoft started things off with a $10 billion partnership in which OpenAI's LLMs are housed on Microsoft Azure cloud servers.

Anthropic was founded in 2021 by former employees of OpenAI, including siblings Dario and Daniela Amodei, and therefore gained instant credibility. The recent $4 billion partnership with Amazon is similar to the OpenAI/Microsoft partnership. Anthropic's LLM and associated chatbot, both named Claude, will be housed on Amazon's AWS cloud.

The AI Arms Race Competition
Facebook/Meta's entry into the AI race is called LLaMA (Large Language Model Meta AI). It was released in February 2023 and has 65 billion parameters. Unlike the OpenAI/Microsoft and Anthropic/Amazon entries, LLaMa is a pure Facebook/Meta effort and is not a partnership.

The Google/DeepMind entry into the AI race is called Gemini. Gemini is still under development, and not much is known about it. The AI community speculates that Gemini will have at least 1 trillion parameters, which will make it perhaps the most powerful LLM on the planet.

[Click on image for larger view.] Figure 1: Images Generated by the DALL-E Tool from Prompt: "AI Race."

A recent startup named xAI has attracted significant attention in the LLM/chatbot world, as it was founded by Elon Musk. The company has only about a dozen employees. In order to create a LLM and associated AI applications, only a few very bright people are needed, along with a very, very fat wallet for computing resources.

What Does It Mean?
So, what does this all mean? One of the Pure AI technical advisors, Dr. James McCaffrey, commented, "To be honest, my colleagues and I fully expected that Anthropic would partner with Amazon. The stakes are obviously huge, and no large tech company, including Amazon, wants to be left behind."

Being a prominent data scientist at Microsoft Research, McCaffrey is in a unique position to observe the AI arms race. "The underlying neural technology of an LLM, while not simple, requires only a relatively small team of engineers," he said. "But housing an LLM requires a huge amount of compute resources, which meant that Anthropic would have to either get billions of dollars in funding, or more likely, partner with a large tech company. Amazon was the obvious choice.

"Because an AI LLM requires so many resources, I don't expect to see a flood of LLM-related startups. But it's an open question about the extent to which government entities will get involved with AI and LLMs."

Looking forward, McCaffrey observed, "Business opportunities are likely to arise where entrepreneurs devise some clever way to use the LLMs developed and maintained by large companies."

The big AI players, however, have gotten a head start in that area, with McCaffrey's employer, Microsoft, already capitalizing on its OpenAI pact by infusing AI-powered assistants dubbed "Copilots" throughout its products and services, recently including even Windows itself.

The rise of Copilots came with accompanying price hikes that, some financial experts say, could lead to an additional $31 billion in revenue.