New Delhi; Amazon Web Services (AWS) has put forth an additional generation of home-built artificial intelligence models and also a service that enables businesses to create highly personalised forms of models. Under its rebranded Nova 2, the company announced four new models as part of its keynote speech by its CEO Matt Garman at the AWS re:Invent on Tuesday. This launch is based on the predecessors of Nova models that were launched last year and is an indication that AWS is seeking to increase its presence in the enterprise-grade AI.
Garman claimed that Nova adoption has significantly increased, and the models are used in tens of thousands by customers in industries. The updates this year are expected to provide more sophisticated reasoning and more flexibility to those organisations that want to have more fine-grained control over AI learning about their own data.
AWS expands Nova model family
The Nova 2 portfolio consists of four products: Nova 2 Lite, an affordable reasoning model used in everyday chores; Nova 2 Pro, a higher-end reasoning model that can perceive text, images, video and speech and produce both text and images; Nova 2 Sonic, a speech-to-speech system that uses conversational AI; and Nova 2 Omni, a multimodal reasoning model that is able to comprehend text, images, video and speech and generate both text and images.
Introducing the Nova forge service
Another service that AWS announced is Nova Forge, in which enterprise customers can design customised Nova family building models, the Nova family, which are called Novella. The service is sold by CNBC for $100,000 annually and provides access to pre-trained, mid-trained or post-trained versions that can be advanced by companies with their own data. According to Garman, this method assists enterprises in escaping problems that are experienced when models lose previous reasoning abilities in the process of overly customising the model, likening it to the challenge of acquiring a second language later in life.