While it already offers Amazon Web Services infrastructures to companies that want to use generative artificial intelligence, Amazon has just presented new foundation models that will be available on the Amazon Bedrock service. Companies thus have a new series of artificial intelligence models optimized for Amazon services, which rival technologies from OpenAI or Google.
Called Nova, these new AI models support 200 languages, and can process texts, images, as well as videos. The 6 models announced by Amazon are:
- Amazon Nova Micro: an artificial intelligence that only supports text, but stands out for low latency and low cost
- Amazon Nova Lite: this model also has a low cost and has a short processing time, but it supports video inputs and images
- Amazon Nova Pro: this model offers a good balance between cost, processing speed and performance
- Amazon Nova Premier: according to Amazon, this is its most powerful model, which will be available in the first quarter of 2025
- Amazon Nova Canvas: a model that allows you to generate images
- Amazon Nova Reel: a video generation model
Amazon also highlights the savings that its customers will be able to make by using its models. “Amazon Nova Micro, Amazon Nova Lite, and Amazon Nova Pro are at least 75% cheaper than the top-performing models in their respective intelligence classes in Amazon Bedrock. They are also the fastest models in their respective intelligence classes in Amazon Bedrock”indicates the tech giant.
Nova Reel: impressive capabilities
Otherwise, Amazon also shared a demonstration of the capabilities of Nova Reel, its video generation AI. One such demonstration is an AI-generated ad for a fictitious brand of pasta.
The company also gives an example of Nova Pro’s ability to understand video. According to Amazon, it sent the video below to artificial intelligence, asking it to generate a description.
And here is the text that Nova Pro produced: “The video shows a football match in progress on a green field. Players from two teams, one in yellow uniform and the other in white, are playing. The yellow team’s quarterback throws a pass to a receiver, who catches the ball and begins running downfield. The white team’s defenders chase him, trying to tackle him. The play ends with a tackle that knocks the receiver to the ground.”
Perfect technologies for Alexa?
And that’s not all, since Amazon is already preparing the release of a speech-to-speech artificial intelligence which will be entirely dedicated to the audio format, as well as a multimodal-to-multimodal technology which will be capable of processing text , images, sound and video, both input and output.
At this time, these announcements only affect Amazon Web Services customers. But as you may already know, Amazon is also working on an improved version of its Alexa assistant. And it is entirely likely that this improved version will rely on Nova AI models.
- Amazon has just presented its artificial intelligence models for Amazon Web Services customers
- The tech giant offers ultra-light models, multimodal AI, as well as image and video generators
- And it is possible that the improved version of Alexa will rely on these new technologies
