AmazonMillions of dollars are being invested in training an ambitiousLarge Language Models, the model was codenamed "Olympus”, hoping to work with OpenAI and AlphabetTopModel competition. The news attracted attention from the outside world, but Amazon declined to comment, according to two people familiar with the matter who told Reuters.
Source: The image is generated by AI, and the image is authorized by Midjourney
According to people familiar with the matter, Olympus has 2 trillion parameters, which would make it the most powerful machine learning algorithm currently in training.maximumOne of the models, and OpenAI's GPT-4 model is consideredoptimalAlthough the details have not been made public, Amazon's goal is clear: it hopes to provide more competitive artificial intelligence services by training "Olympus".
This training teamLeadersIt was Rohit Prasad, who previously served as the head of Alexa and now reports directly to Amazon CEO Andy Jassy. As Amazon's chief AI scientist, Prasad brought together researchers who had worked on Alexa AI and Amazon's scientific team to work together on training models.
Amazon has trained some smaller models, such as "Titan," and has also used them toAI ModelsStartups such as Anthropic and AI21Labs have collaborated to make these models available to users of Amazon Web Services (AWS). Amazon believes that having its own models can make its services on AWS more attractive because corporate customers want access to models that perform well.
Large language models (LLMs) are the foundational technology for AI tools that support learning from large data sets to generate human-style responses. However, training larger AI models is more expensive because a lot of computing power is required. In an earnings call in April this year, Amazon executives said the company would increase investment in LLMs and generative AI while reducing fulfillment and transportation in its retail business.