Reports indicate that Amazon will launch a large-scale natural language model “Olympus”
According to information acquired by the website, Amazon is poised to unveil a colossal natural language model, boasting an astounding 2 trillion parameters and codenamed ‘Olympus’, anticipated to launch as early as this December.
In contrast to OpenAI’s GPT-4, which encompasses 1 trillion parameters, Amazon’s planned ‘Olympus’ natural language model will double in parameter size. It is expected to debut during the upcoming re:Invent 2023 conference and is slated to enhance its e-commerce services, Echo, and other smart devices. It is also anticipated to be integrated into the AWS cloud platform, thereby expanding accessibility for a multitude of developers and enterprises.
Previously, Dave Limp, the Senior Vice President of Amazon Alexa and Devices, mentioned that the future of the Alexa digital assistant service will incorporate generative artificial intelligence, enabling users to interact in a manner that more closely mirrors natural conversation and assists them in accomplishing a broader array of tasks. At that time, Limp hinted that Alexa might evolve into an independent, subscription-based service due to the substantial operational costs of the underlying models.
Artificial intelligence currently stands as a focal point of Amazon’s development initiatives. CEO Andy Jassy has expressed that AI is set to revolutionize numerous business models. Recently, Amazon has been actively investing in AI applications, including a significant $4 billion investment in Anthropic, allowing it to integrate its large-scale natural language model with AWS services, and granting AWS customers the ability to craft a diverse range of applications with this advanced language model.