Amazon Web Services (AWS) has launched new AI models and a service aimed at giving enterprise customers more control over their AI capabilities. Key points include:
-
Launch of Nova 2 Models: During the AWS re:Invent keynote, AWS CEO Matt Garman introduced Nova 2, which includes four upgraded AI models:
- Nova 2 Lite: A cost-effective reasoning model for everyday tasks.
- Nova 2 Pro: A reasoning agent designed for complex tasks, such as coding, capable of processing various data types.
- Nova 2 Sonic: A speech-to-speech model intended for conversational AI applications.
- Nova 2 Omni: A multimodal model that can handle images, text, video, and speech, producing both text and images.
-
Introduction of Nova Forge: AWS also announced Nova Forge, a service allowing customers to create custom versions of Nova models, termed Novellas, for an annual fee of $100,000. This service provides access to pre-trained, mid-trained, or post-trained models that enterprises can further train with their proprietary data.
-
Customer Adoption: Garman highlighted that Nova has gained traction, being utilized by tens of thousands of customers, including major companies like Infosys, Blue Origin, and Robinhood, as well as innovative startups.
-
Challenges in Model Customization: Garman noted that customizing AI models can lead to challenges, where models may "forget" core reasoning if too much new data is added post-training, likening it to the difficulties humans face when learning new languages later in life.
The developments indicate AWS's commitment to enhancing AI capabilities for businesses, enabling greater customization and control over AI applications.
