Amazon Develops AI Model “Olympus” to Compete with ChatGPT and Bard

Amazon is racing to create a cutting-edge AI model codenamed “Olympus” to stay competitive in the evolving AI landscape and challenge rivals like OpenAI and Google. The tech and commerce behemoth is aiming to seamlessly integrate this large language model into its online store and Alexa smart speakers. According to a report by The Information, “Olympus” could potentially be unveiled as early as December, signifying Amazon’s determined entry into the AI arms race. This move follows Amazon’s significant $4 billion investment in AI startup Anthropic, announced in September, as the company makes AI a top priority in its pursuit of creating products that can rival the capabilities of platforms like ChatGPT and Bard.

A Glimpse into “Olympus”

“Olympus” is poised to become one of the largest AI models ever constructed, featuring an impressive two trillion parameters compared to GPT-4’s one trillion. However, experts caution that size alone does not guarantee superior performance in AI models. Amazon’s previous AI model, Titan, faced a delayed launch last year due to technical complications and the introduction of ChatGPT, which Amazon executives perceived as surpassing their offering.

Closing the Gap

In response to these developments, Amazon is diligently working to bridge the gap. The company is actively developing a “conversational shopping agent” named “Project Nile” for its online store, while also announcing plans in September to integrate generative AI into its Alexa smart speakers. These AI enhancements, rooted in a custom large language model, will empower Alexa to engage in conversations with users, compose messages, and respond to more intricate queries. Although Alexa’s AI tools are currently undergoing a preview phase with select users, Amazon has hinted at the possibility of introducing a subscription fee for access to these advanced features in the future.

Leave a Reply

Your email address will not be published. Required fields are marked *