Image: SOPA Images / Contributor.
If Amazon has massively adopted generative AI on its various platforms and developed its use, the e-commerce giant has not yet unveiled its own competitor to ChatGPT.
But that could change, according to a survey that suggests the company is investing millions in a new large language model (LLM).
Olympus Project
Last Wednesday, The Information published an article revealing that Amazon is reportedly building its own conversational LLM, codenamed “Olympus.” The source of the media is a person who has direct knowledge of the information.
According to information in the article, Olympus is being developed to be sold to businesses, like the solutions offered by OpenAI and Microsoft, respectively ChatGPT Enterprise and Microsoft Copliot.
Amazon’s new LLM should be used to power new features both on the popular e-commerce site, but also on Alexa voice assistance – on Echo devices – and on Amazon Web Services.
Amazon could announce Olympus at its next event, AWS re:Invent 2023. It is expected to perform better than Titan, the LLM currently sold by AWS to its customers, and which is lower in capacity than its competitors, notably GPT-4 from OpenAI.
The most advanced LLM, twice the size of GPT-4
Concerning the details of the project, Reuters learned, from two sources close to the matter, that Olympus would have two trillion parameters.
For comparison, GPT-4, OpenAI’s most advanced LLM, has a trillion parameters. So, if the information is confirmed, Amazon’s next model would be twice as large as the market leader, potentially becoming one of the largest LLMs ever built, the article suggests.
As early as May, ZDNet reported news that Amazon would build its own AI-powered chatbot. These rumors were until now mainly based on job offers, published by Amazon, and reporting this initiative.
Source: ZDNet.com