Amazon is working on its own ChatGPT competitor. Meet project Olympus

Inside sources reveal that Amazon is gearing up to release the most advanced LLM to date, according to reports.
Written by Sabrina Ortiz, Editor
Amazon and AI logos
SOPA Images / Contributor

While Amazon has not shied away from developing and adopting generative AI across its different platforms, the company is yet to unveil its own rival to ChatGPT. However, a new report suggests the company is investing millions into a new large language model (LLM). 

On Wednesday, The Information published a report that revealed Amazon is building its own conversational LLM, codenamed 'Olympus'. The report was informed by a person with direct knowledge. 

Also: Watch out: Generative AI will level up cyber attacks, according to new Google report

Olympus is being developed to be sold to corporate customers, much like the enterprise solutions that OpenAI and Microsoft offer, such as ChatGPT Enterprise or Microsoft Copliot, according to the report. 

Amazon's new LLM will supposedly be used to power new features in its online retail store, Alexa voice assistant across its Echo devices, and its Amazon Web Services unit, according to the insider. 

Olympus could be announced as soon as the upcoming AWS re:Invent 2023 event and is slated to perform better than Titan, a group of LLMs that AWS is currently selling to cloud customers and whose capacity is inferior to competitors, such as OpenAI's GPT-4

According to a Reuters report informed by two people familiar with the matter, Olympus will have two trillion parameters. 

Also: Google's Performance Max gets new AI tools to help automate ad creation even further

For context, GPT-4, OpenAI's most advanced LLM, has one one trillion parameters, making Amazon's model twice as big and potentially one of the largest LLMs ever built, according to the report

ZDNET reported rumors of Amazon building its own AI chatbot as early as May, which was based on job listings that Amazon posted that pointed to an AI chatbot-building initiative. 

Editorial standards