Amazon unveils Nova, new set of AI models

DailyStar || Shining BD

Published: 12/5/2024 6:38:50 AM
A screenshot from an example video generated by Amazon Nova, with the prompt: 'A teddy bear in a leather jacket, baseball cap, and sunglasses playing guitar in front of a waterfall'. Image: Amazon Web Services official YouTube channel

A screenshot from an example video generated by Amazon Nova, with the prompt: 'A teddy bear in a leather jacket, baseball cap, and sunglasses playing guitar in front of a waterfall'. Image: Amazon Web Services official YouTube channel

Amazon has introduced Nova, a collection of advanced AI models designed for tasks like analysing text, generating images, and creating videos. 

These models, announced at the AWS conference in Las Vegas on Tuesday, aim to help businesses reduce costs, speed up processes, and customise AI for their specific needs, says Amazon in an official blog post.

Nova is available through Amazon Bedrock, Amazon's platform for integrating AI into business applications. The models are divided into two primary categories: 'understanding' and 'creative' content generation.

'Understanding' models

The 'understanding' models process text, images, and videos to produce text-based outputs.

Among these, the Amazon Nova Micro model focuses on text-only tasks, such as summarisation, translation, and simple coding, with an emphasis on low latency and cost efficiency.

Another model, Amazon Nova Lite, can handle text, image, and video inputs, analysing up to 30 minutes of video or multiple images in a single request.

Amazon Nova Pro combines speed, accuracy, and cost efficiency to handle more complex tasks, such as processing financial documents or analysing large codebases, says Amazon.

The company has also announced Amazon Nova Premier, a model still in development, which is expected to handle more advanced reasoning tasks and provide additional customisation options upon its release in 2025.

'Creative' models

Nova also includes 'creative' models for producing visual content. According to Amazon, Amazon Nova Canvas creates high-quality images from text prompts and offers features like background removal and image editing.

The other model, Amazon Nova Reel, generates short videos using text or images. Currently, it supports six-second videos, with plans for longer videos soon.

Both models include safety features like watermarking to prevent misuse.

Customisation options and future plans

Nova models allow businesses to fine-tune them with their own data. This customisation can help the models understand specific industry terms or adapt outputs to match a company's style or goals, says Amazon. For example, a law firm could train the models to understand legal documents better.

Amazon's Nova launch comes as the company competes with other tech giants like Adobe, Meta, and OpenAI in the rapidly evolving AI space.

As per a Reuters article, CEO Andy Jassy explained during the unveiling that developers wanted faster, cheaper, and more flexible tools, prompting the creation of Nova. Rohit Prasad, Amazon's head of AI, emphasised that Nova's speed and cost advantages could help it stand out in the market.

Amazon also plans to enhance its Alexa voice assistant with advanced AI, as part of an initiative called Banyan. While the project has faced delays, the company remains optimistic about improving voice assistant technology, states official sources.

Shining BD