What is Amazon Bedrock? The custom GenAI building service

It’s painfully cliché these days to start an article with “there’s a lot of hype around generative AI right now” — even if you’re living under a rock, someone has no doubt walked past it and murmured something about ChatGPT. What is less obvious, though, is how to actually apply it in your organization or inside your apps.

Part of the problem is that the out-of-the-box solutions are 70% good at doing what 70% of people want them to do. But what if your use case is more bespoke, or you want more reliable output from an AI model? You want something custom and trained on your data, something nobody is likely to sell.

In that case, knowing a lot about rocks is actually useful: specifically, Amazon Bedrock.

What is Amazon Bedrock?

Amazon Bedrock is a fully managed service for building generative AI applications. You start with a foundational model (FM) from either Amazon or a third-party AI company through an AI, then build up from the pre-trained “bedrock” of the base model to customize it to your needs. 

Because it’s fully managed, you don’t have to manage the infrastructure at all, and you can privately customize your FM using your organization’s data. There’s a wide range of supported FM providers to choose from, too, such as:

• Amazon

• Anthropic

• AI21 Labs

• Cohere

• Stability AI

What can I use Amazon Bedrock for?

Since you can use Amazon Bedrock to build and scale generative AI models, you can do pretty much everything you can do with GenAI in general, such as:

1. Building chatbots

You can create and customize chatbots with Bedrock, then add them to your website or a stand-alone section. 

2. Text and image generation

The bread and butter of generative AI models. You can generate a wide range of original written or image-based content from language prompts.

3. Searching through data

There is a search function that allows you to search through information to answer questions from huge amounts of data.

4. Text summarization

Text summarization is also a build option. This feature allows you to get a quick and easy summary of any text based content, such as articles, blog posts, books, or documentation.

How do I get access to Amazon Bedrock?

Right now, Amazon Bedrock is invite only, so it’s not generally available to the public. You need to submit a request to have access to that side of the console. There is a form you can fill out on the product page, but unfortunately that only leads to a generative AI newsletter sign-up form.

How much does Amazon Bedrock cost?

Because the service is in development and invite only, the pricing is not set in stone. It varies wildly depending on what you’re building.

What’s the difference between Amazon Bedrock and SageMaker?

Amazon Bedrock is a fully managed service which makes it easy to build and scale generative AI applications with foundational models, and allows for light customization. In comparison, Sagemaker is a managed ML service that allows for deep customizability and is for more advanced use-cases.

Getting more into the details, SageMaker is a way to train, build, and deploy models, whereas Bedrock deals with pre-trained models. With SageMaker, you can use any available open-source Large Language Model (LLM) and train it. In comparison, Bedrock is limited to default Foundational Language Models, even though you can fine-tune them.

Conclusion

I hope this article has cleared some things up about what Amazon Bedrock is, and where it can be useful. Keep your eyes peeled for when Amazon Bedrock becomes more widely available. In the meantime, if you’re interested in other Generative AI services offered by Amazon, check out my A Cloud Guru course: “Foundations for Generative AI in AWS”.

Pluralsight also offers a range of beginner, intermediate, and expert AI and ML courses, including dedicated courses on generative AI and tools like ChatGPT. Since you can sign up for a 10-day free trial with no commitments, it’s a great way to take some professionally authored courses with a set course structure.