AWS has joined the generative AI march. But, instead of building models entirely by itself, it is recruiting third parties to host models on its cloud.
Enter Amazon Bedrock. It provides the platform for users to build generative AI-powered apps from pre-trained models created by various startups. These include AI21 Labs, Anthropic, and Stability AI.
In doing so, AWS is giving the pick axes, shovels, and drills to miners instead of digging for the treasures itself – as Microsoft and Google are doing.
Such a strategy is not new for AWS, according to Andy Jassy, CEO of Amazon. He told CNBC:
If you think about what AWS has been doing for many years, we’ve been really trying to democratize technology so that large and small companies alike can afford and have access to build amazing customer experiences.
Until now, organizations that wished to build custom large language models (LLMs) for their unique business requirements would likely have to train the system for many years. In addition, they’d spend eye-watering sums of money.
That does not sound like an appealing prospect. AWS believes it can offer a differentiator by offering foundational models, which businesses can customize for their unique purposes.
This mission sits at the core of Bedrock, which serves up many of these models from third parties – alongside its own "Titan" models.
Now, these are not consumer models like ChatGPT or Bard. Instead, they developer-focused LLMs, which businesses can finetune and build generative AI experiences on top of.
To support developers in this mission, AWS has also launched "CodeWhisperer."
This generative AI application allows a developer to write out – in natural language – what they want to build, and it writes the code for them.
ChatGPT offers a similar function, which will help to transform developer productivity.
