Nov 29, 2024 12:06:00

Amazon's new generative AI model, Olympus, can process images, videos, and text and could be unveiled as early as next week at AWS re:Invent

Amazon is developing a multimodal AI model called 'Olympus' that can process images, videos, and text, and is likely to announce it at

AWS re:Invent, which is scheduled to be held from December 2 to 6, 2024, according to the American news site 'The Information.'

Amazon Develops Video AI Model, Hedging Its Reliance on Anthropic — The Information
https://www.theinformation.com/articles/amazon-develops-video-ai-model-hedging-its-reliance-on-anthropic

Amazon to launch 'Olympus': A new generative AI model for video, image, and text processing - Technology News | The Financial Express

https://www.financialexpress.com/life/technology-amazon-to-launch-olympus-a-new-generative-ai-model-for-video-image-and-text-processing-3679253/

Olympus understands scenes in images and videos, so it can search for specific scenes based on user text prompts, such as 'the winning shot in a basketball game' or 'the cooking scene.'

Amazon has invested $4 billion (approximately 600 billion yen) in AI company Anthropic so far, and has announced that it will invest an additional $4 billion (approximately 600 billion yen) on November 22, 2024. In addition to investing in Anthropic, Amazon has built a partnership by introducing Anthropic models such as Claude to AWS's generation AI service 'Bedrock'.

Amazon invests $4 billion into Claude developer Anthropic, co-developing AWS's own AI chip - GIGAZINE

Meanwhile, Olympus is being developed independently by Amazon, with The Information speculating that the aim is to reduce reliance on Anthropic.

Amazon has introduced its own generative AI such as Amazon Titan to Bedrock, and Olympus is expected to offer it as a managed service through Bedrock. Users can use advanced multimodal AI models without the need for large-scale infrastructure investment, and it is expected to be used in fields that require the integration of image analysis and natural language processing.

According to a Reuters report in November 2023, Olympus is developing a 2 trillion parameter model. Development has not been going smoothly, and in March 2024, it was reported that the company was 'struggling to surpass the performance of competitors' models.'