Amazon is preparing a competitor to OpenAI: the team is working on its own Olympus AI model
Online retailer Amazon has formed a team to train Olympus, a large language model of artificial intelligence, as the company seeks to compete with OpenAI and Alphabet.
It is stated that the company is investing millions in training a large language model (LLM), which is the underlying technology for AI tools. It is trained on huge data sets so that it can generate human-like responses.
Amazon believes that having its own model will make the company more attractive to Amazon Web Services (AWS) users. It is a subsidiary of Amazon.com that provides a cloud computing platform for rent to individuals, companies, and governments on a paid subscription basis. Its clients seek the most efficient AI model. So Amazon decided to increase funding for the project by reducing order fulfillment and transportation costs in its retail business.
According to the publication’s sources, the model, codenamed Olympus, has 2 trillion parameters, which may mean that it is currently the largest learning model. OpenAI’s GPT-4 model reportedly has one trillion parameters.
The Olympus training team is led by Rohit Prasad, who previously led the work on the Alexa voice assistant. He reports directly to CEO Andy Jassy.
“As chief scientist of artificial general intelligence (AGI) at Amazon, Prasad has engaged researchers who worked on Alexa AI and Amazon’s science team in the development of training models, bringing together AI efforts across the company with dedicated resources,” the statement said.
The sources indicated that the timing of the Olympus release is currently unknown.
Before that, Amazon trained smaller models, such as Titan. The company also cooperates with AI startups (Anthropic and AI21 Labs). It offers their services to AWS users.
- Earlier, Amazon launched the first Project Kuiper Internet satellites The company hopes to compete with Elon Musk’s SpaceX.
