|Amazon Working On Olympus LLM
|Written by Kay Ewbank
|Friday, 24 November 2023
According to Reuters, Amazon is working on a large language model (LLM), codenamed "Olympus". With, reportedly, 2 trillion parameters, it is twice the size of OpenAI's GPT-4 model.
Since Reuters carried news of Olympus, which was based on information from two Amazon insiders familiar with the project but wanting to keep their anonymity, there has been no confirmation nor denial from Amazon. However it seems highly likely that Amazon is intending to produce an LLM to go up against those of OpenAI and Google.
LLMs are currently at the forefront of advances in artificial intelligence. Given their ability to generate text, translate languages, write different kinds of creative content, and answer questions in an informative way they have quickly gained popularity by end users. They are trained on massive amounts of data but if the model for the Olympus project is very large with 2 trillion parameters, as claimed by Reuters' sources, this would make it the largest models currently being trained, twice as large as OpenAI's GPT-4 model, on which ChatGTP is based, which has one trillion parameters and nearly four times are large as Google's PaLM, which powers Bard. The larger the model, the greater the associated costs because of the amount of computing power required.
Amazon has already trained smaller models such as Titan, a family of foundation models pretrained by AWS on large datasets, making them powerful, general purpose models. Amazon has also announced an expanded model choice with Llama 2 (coming in the next few weeks) and Amazon Titan Embeddings. Amazon has also worked with other companies creating AI models including Anthropic and AI21 Labs, offering them to Amazon Web Services (AWS) users.
Reuters said the people it spoke to says that Amazon believes having homegrown models could make its offerings more attractive on AWS, where enterprise clients want to access top-performing models. There is apparently no specific timeline for releasing the new model - although there is a rumor, reported by Bard and therefore subject to suspicion, that it could be as soon as December 2023.
The news of the new model follows a statement by Amazon executives in April in which spokespeople said Amazon would increase investment in LLMs and generative AI while cutting back on fulfillment and transportation in its retail business.
According to Reuters, the team working on Olympus is being led by Rohit Prasad, who used to be head of Alexa at Amazon. Prasad now reports directly to Amazon's CEO, Andy Jassy. Other members of the Olympus team are, again according to Reuters, researchers who have worked on Alexa AI. Others have been recruited from the Amazon science team.
or email your comment to: firstname.lastname@example.org
|Last Updated ( Friday, 24 November 2023 )