Amazon’s Bedrock AI

In mid-April, Amazon announced its entrants into the generative AI race: Bedrock, a platform for using different foundational models (FMs) for generative AI within AWS, and Titan, Amazon’s own large language model (LLM) for usage in a variety of language-related tasks. Amazon is partnership with Stability AI, Anthropic, and A21I Labs to offer access to their generative AI models.

While the reaction in the business press was quite positive, on further reflection it seems this announcement suggests Amazon is behind in the AI race, at least with regard to generative AI. OpenAI (and be extension Microsoft Azure) is the clear leader with GPT-4. Google went to “code red” internally in reaction to OpenAI and put out its Bard LLM in March as well, although the performance of the product seemed 1-2 years behind GPT-4. To its credit, Google updated Bard a month later with code generation and debugging, showing an ability to ship product that should keep it in the race. Amazon in contrast has mostly been silent and now comes out with an announcement with surprisingly little meat to it.

To be fair, Amazon does clearly state that Bedrock and Titan are in preview. From what I’ve seen, there is currently only one customer, Coda, presented as an adopter. It could well be that there is a lot of customer adoption and thinking about how to improve AI workflows behind the scenes, but at this point the offering feels a bit like a placeholder.