Amazon Bedrock, a new service that makes FMs from AI21 Labs, Anthropic, Stability AI and Amazon accessible via an API
amazon announced recently the launch of its artificial intelligence platform called “Bedrolk”, which is a set of tools for allow AWS customers to create chatbots and generate text, In addition, it can allow users to create and classify images. Amazon Bedrock provides APIs to access generative AI models created by AI21 Labs, Anthropic, and Stability AI.
The aim of the launch is to target legal entities that want to integrate AI into their business, as Amazon's newly launched platform will host leading language models from startups such as AI21 Labs, Anthropic and Stability AI.
In fact, as Meta, Microsoft, and Google all race to introduce AI-generative chatbots into search and productivity suites, Amazon's strategy is to remain fairly neutral and provide access to the latest models on its cloud-based platform.
“Customers have told us that there are some major roadblocks holding them back today. First, they need an easy way to find and access high-performance entry-level models that perform exceptionally well and best suit their needs. Second, customers want seamless application integration without having to manage large infrastructure clusters or incur significant costs,” said Swami Sivasubramanian, Vice President of Machine Learning at AWS.
Amazon Bedrock, will allow to Amazon Web Services (AWS) users develop a generative AI based on Foundation Models (FM). Bedrock is a serverless experience where customers can privately customize FMs with their own data and easily integrate and deploy them into their applications. The partnership between Amazon and Anthropic and Stability AI will allow Bedrock users to have access to major model languages such as Claude and Stable Diffusion.
Bedrock's suite of text-to-image AI models hosted by Stability AI, including the popular Stable Diffusion, can generate images, art, logos, and graphic designs. Besides, Bedrock is also home to AI2 Labs' Jurassic-21 family of AI models, which is multilingual. and can generate text in Spanish, French, German, Portuguese, Italian and Dutch. Although Bedrock focuses on third-party models, the company features foundational models developed in-house.
For example, Bedrock will also house Titan, a family of two AI models, Similar to OpenAI's GPT-4 model (but not necessarily at the same level of performance), Titan can perform tasks like writing blog posts and emails, summarizing documents, and extracting information from databases. Titan is a model of integration. This means that it translates textual inputs, such as words and sentences, into numerical representations, known as embeddings, which contain the semantic meaning of the text. Not to be confused with Google's AI Titan model.
On the other hand, it is worth mentioning that Amazon clarifies that the data they add when using a hosted model in Bedrock will not be integrated into the same, so these data will remain private and protected from other potential clients and big rivals in the sector, since due to the recent problems that ChatGPT went through, companies fear that the data they share with AI models will be memorized and then leaked.
The costs of training an AI model can be quite high, so the various AI solutions released often use a model that is trained on a given set of data, which essentially gets "dirty" with that data and is prone to leaking. information from them in response to unrelated inquiries.
That is why Bedrock allows customers to circumvent this problem by giving them the ability to use legacy FMs as the backbone to back up their data. With Bedrock, Amazon is sidelining significant investments in one particular AI startup. The company says it wants to democratize AI by making it easier to access AI models.
finally if you are interested in knowing more about ityou can refer to the original post In the following link.