Amazon Invests Big Funds to Train AI Models, for OpenAI and Alphabet

Amazon invests in AI model with code "Olympus", (photo: twitter @awscloud)

JAKARTA - Amazon is throwing millions of dollars into ambitious large language model (LLM) training. They hope to rival top AI models from OpenAI and Alphabet.

The model, known by the code name "Olympus", has 2 trillion parameters, two of the sources said. This AI model could make it one of the largest models being drilled. OpenAI's GPT-4 model, one of the best models available, reportedly has one trillion parameters.

The sources spoke on condition of anonymity because details of the project have not been made public. Meanwhile Amazon declined to comment. Information reported the name of this project on Tuesday, November 7.

The team is led by Rohit Prasad, former head of Alexa, who now reports directly to CEO Andy Jassy. As chief artificial intelligence (AGI) scientist at Amazon, Prasad brought in researchers who had previously worked on Alexa artificial intelligence and Amazon's science team to work on training models, unifying artificial intelligence efforts across the company with existing resources.

Amazon is already training smaller models like Titan. The company has also collaborated with artificial intelligence startup models such as Anthropic and AI21 Labs, which offer such services to Amazon Web Services (AWS) users.

Amazon believes that having its own homegrown models could make its offering more attractive on AWS, where enterprise clients want access to the best-performing models. The source  also added that there is no specific timeline for the release of this new model.

LLMs are the underlying technology for artificial intelligence tools that learn from large data sets to generate human-like responses.

Training larger artificial intelligence models is, of course, more expensive considering the amount of computing power required. In an earnings call in April, Amazon executives said that the company would increase investments in LLMs and generative artificial intelligence while reducing spending on fulfillment and transportation in its retail business.