Amazon has decided to enter the race of the main technology companies to take positions in the field of generative Artificial Intelligence. As proof of this, it has announced AWS Bedrocka new APIs for Amazon Web Services that allows developers use and customize Artificial Intelligence tools capable of generating text or images. In addition, it has announced that its development support AI tool, codewhispererbe free for individual developers.
AWS Bedrock is intended as a configurable, cloud-based alternative that can serve as an alternative to DALL-E2 or ChatGPT. Designed for companies and developers, it allows AWS customers to develop and deploy chatbots, summarize texts, classify images and other tasks, based on messages and text requests.
Bedrock users have the possibility to choose between several models for their tools. Among them are Amazon’s foundational models, Titan Text and Text Embeddings. The mission of the first is to generate texts from requests made also through texts, while the second is to generate a mathematical representation of text that can be used in various tasks. Among them, for translation or search.
But these are not going to be the only models that AWS Bedrock has access to. It will also stop working with other startups dedicated to AI. Like Claude from Anthropic, Jurassic-2 from AI21 (language model specialized in Spanish, French, German, Portuguese, Italian and Dutch) or Stable Diffusion, from Stability AI.
In addition, companies and developers can, with Bedrock, customize the work of models based on the inputs. Of course, according to Amazon, these inputs will not be used to train the models, a note that addresses one of the main concerns of companies and users of these AI services about data privacy.
The fact that Amazon offers various Artificial Intelligence models through AWS Bedrock has to do with offering flexibility to customers, on which the company highlights that “with Bedrock’s serverless experience you can start quickly, privately customize models foundational data with your own data and easily integrate and deploy it into your applications, using the AWS tools and features you’re familiar with. Including integrations with Amazon Sage Maker ML features, such as Experiments, to test different models, and Pipelines, to manage your foundational models at scale, without having to manage any type of infrastructure.
With this service, Amazon offers customers who need it everything they need to work with a foundational model and be able to customize it to cover their needs. His early clients include C3.ai, Pegasystems, Accenture and Deloitte. For now, the price plans that AWS will launch to use the service, which currently has controlled access through a waiting list on its website, are unknown.
Regarding the launch of the definitive version of CodeWhisperer, available as a trial since last year, can be used in various integrated development environments, or IDEs. Among them is Visual Studio Code, to generate lines of code based on the text requests it receives. Initially it was intended for AWS customers, but from now on it will have a free version for which it will not be necessary to use Amazon cloud services.
Among other functions, CodeWhisperer is responsible for filtering any code suggestions that may contain some kind of bias, or that are not fair. It also tags any code that is similar to the open source training data. In addition, it incorporates security scanning capabilities, with which it can identify vulnerabilities in code written by a developer, as well as offer suggestions to help close security issues it identifies.
CodeWhisperer supports several languages, including some of the most popular. Among them Python, PHP, Java, JavaScript, TypeScript, C, C++, C#, Go, Rust, Ruby, Kotlin, SQL or Scala.
On the other hand, everything points to the fact that Amazon does not intend to focus solely on these services and tools in its commitment to AI. In addition, in the company they want their shareholders to be aware of this. That’s why its CEO, Andy Jassyhas sent them a letter to inform them, among other things, that at Amazon they are investing heavily in large scale language modelsas well as in Generative Artificial Intelligence.
In the letter, Jassy assures that they have been working on Amazon’s own large language models for some time, and that they believe that with this they will be able to transform and improve the experience of practically all customers. Also that they will continue to invest in these models to improve the experiences of consumers, sellers, brands and creators.
Jassy also points out that Amazon’s goals include offer less expensive machine learning chips, so that both large and small companies can afford to train and run their large language models in production. Also remember that many companies want to use these models, but the really good ones require investments of billions of dollars in training.
Apart from this, they can take many years to get them ready, something that most companies cannot afford. That is why they have launched Bedrock, a service that Amazon believes will change the landscape in terms of AI for many companies and organizations.
