Amazon is making a substantial investment in training a formidable large language model (LLM) known as ‘Olympus,’ with hopes of competing with leading models developed by OpenAI and Alphabet, according to sources familiar with the matter. With a staggering 2 trillion parameters, Olympus could potentially become one of the largest models ever trained, surpassing OpenAI’s GPT-4, which boasts 1 trillion parameters.
Amazon’s move into the realm of advanced AI models reflects its commitment to remain at the forefront of AI innovation. While the details of the project have not yet been publicly disclosed, sources indicate that Amazon is dedicating significant resources to the development and training of Olympus.
Leadership by Rohit Prasad
Rohit Prasad, former head of Alexa, is spearheading the team responsible for Olympus. Prasad, who now reports directly to Amazon CEO Andy Jassy, serves as the head scientist of general artificial intelligence (AGI) at Amazon. Under his leadership, researchers who previously worked on Alexa AI and the Amazon science team have been brought together to collaborate on training models. This strategic consolidation of AI efforts within the company demonstrates Amazon’s commitment to advancing AI technologies.
Expanding the AI ecosystem
Amazon’s foray into large language models is not limited to Olympus. The company has also ventured into training smaller models, including Titan. Additionally, Amazon has formed partnerships with AI model startups such as Anthropic and AI21 Labs, offering these models to users of Amazon Web Services (AWS). This expansion of its AI ecosystem is seen as a strategic move to enhance AWS’s appeal to enterprise clients seeking access to high-performing AI models.*
Amazon’s investment in developing homegrown AI models aligns with its broader strategy to make AWS more attractive to enterprise clients. These clients increasingly require access to top-performing AI models for their applications. By harnessing the power of Olympus and other AI models, Amazon aims to establish a competitive edge in the cloud computing space.*
Challenges in training large AI models
*Training larger AI models comes with significant challenges, primarily due to the immense computing power required. This undertaking incurs substantial expenses, but Amazon is willing to make this investment to advance its capabilities in large language models and generative AI.
Amazon’s focus on bolstering its AI capabilities signifies a strategic shift in its investment priorities. In an earnings call held in April, Amazon executives announced their intention to increase investments in LLMs and generative AI while reducing investments in fulfillment and transportation within the retail sector.
As of now, there is no specific timeline for the release of Olympus or when it will be made available to the public. However, Amazon’s dedication to this ambitious AI model underscores the company’s commitment to remaining competitive and innovative in the AI landscape.
Amazon’s dedication to training the Olympus large language model represents a significant step in its quest to compete with industry leaders in the field of AI. With Olympus’ massive parameter count, Amazon aims to position itself as a major player in the AI model space, particularly within its cloud computing offerings through AWS. The company’s strategic investment in AI, under the leadership of Rohit Prasad, demonstrates its determination to stay at the forefront of AI and generative AI technologies. As Amazon continues to develop and refine its AI capabilities, the technology industry eagerly awaits the unveiling of Olympus and its potential impact on the AI landscape.
