Amazon Web Services launched Amazon Nova, a series of foundation models available in Bedrock, in a move that aims to provide large language model choice and commoditize the market.
Think of Amazon Nova as the Trainium and Inferentia strategy applied to genAI models. AWS is betting that enterprises will follow the money and opt for Amazon Nova on Trainium with the Bedrock stack.
The models include Amazon Nova Micro, Nova Lite, Nova Pro and Nova Premier with additional models on deck. While that's the news, it's worth thinking through the big picture of how AWS is approaching models.
- AWS aims to make Amazon Bedrock your agentic AI point guard
- AWS revamps S3, databases with eye on AI, analytics workloads
- AWS scales up Trainium2 with UltraServer, touts Apple, Anthropic as customers
- AWS outlines new data center, server, cooling designs for AI workloads
- AWS re:Invent 2024: Four AWS customer vignettes with Merck, Capital One, Sprinkr, Goldman Sachs
- Oracle Database@AWS hits limited preview
AWS' bet is that LLMs will be a commodity that will be mixed and matched depending on the task at hand. Speaking during the AWS re:Invent 2024 keynote, Jassy said that Nova will be tightly integrated with AWS services to deliver lower latency and price performance.
Jassy said during the re:Invent 2024 keynote that the company is focused on choice and noted that there will be multiple models used in applications. Jassy noted that the Alexa rebuild will use multiple models. Jassy said:
"We are learning the same lesson over and over and over again, which is that there is never going to be one tool to rule the world. It's not the case in databases. It's not the case in analytics. We were talking about how everybody thought the TensorFlow was going to be the one AI framework. There were a lot of them and Pytorch ended up being the most popular one. The same is going to be true for models. Our internal builders have been asking for all sorts of things from our teams that are building models. They want better latency. They want lower cost. They want the ability to fine tuning. They want the ability to better orchestrate across their different knowledge bases, to be able to ground their data. They want to take lots of automated, orchestrated actions, or what people call agentic behavior. They want a bunch."
Key points:
- Amazon Nova will add speech-to-speech and any-to-any models coming soon.
- Amazon Nova Canvas will focus on image generation and Reels will generate video.
- Nova aims to be 75% more cost effective.
- Integrated into Bedrock, support fine tuning and be optimized for agentic AI.
Jassy concluded:
"We always provide you selection everything we do, which is that we are going to give you the broadest and best functionality you can find anywhere. It's going to mean choice. You are going to use different models for different reasons at different times, which is the way the real world works. Human beings don't go to one human being for expertise in every single area. You have different human beings who are great at different things."
Constellation Research analyst Holger Mueller said:
"AWS reverts its position on LLMs and gets in the market with its Nova models. It's a sign that Amazon / AWS realize they need LLM offerings for both in-house and customer use cases. This will temporarily affect its 'Switzerland' of AI position. Its strong appeal was for LLM vendors was to partner with Bedrock whilst there was no in-house LLM competition. But AWS knows how to partner."
