Amazon studies Olympus: details on the artificial intelligence model that aims to surpass ChatGPT

It would appear that Amazon is investing large sums of money in training an ambitious large language model (LLM)which can compete with OpenAI e Alphabet. The news, revealed by two informants a Reutershighlights that the model would have “Olympus” as a code name and that you can count on 2000 billion different parameters: therefore, it could turn out to be one of the largest models in the training phase. In this regard, making a quick comparison, OpenAI’s GPT-4 model is based on a trillion parameters.

Specifically, the team would be led by Rohit Prasad, former head of Alexa, who will have to report news and developments of the project directly to the CEO Andy Jassy. As chief scientist of artificial general intelligence (AGI) at Amazon, Prasad it would involve researchers who had worked on Alexa AI and Amazon’s scientific team to work on training models, thus uniting the efforts of the entire company in the field of artificial intelligence (also being able to count on dedicated resources).

In the past, Amazon has already trained smaller models, such as Titan. Furthermore, it has collaborated with startups in the field such as Anthropic and AI21 Labs, for the purpose of offering additional services to users of Amazon Web Services (AWS). According to Reuters, however, the company was convinced to develop internal models because it was convinced that this could make its offers on AWS more attractive. As for the release of Olympus, certain times have not yet been communicated.