The public is now also introduced to the company’s AI image generator. Along with the newly launched AI chatbot Amazon Q, it is clear that the company wants to keep up with the competition in the generative AI field. To show that it keeps abreast of what is going on in the field, it is adding the latest foundation models to Amazon Bedrock.
The Titan Image Generator is OpenAI’s AI image generator. The company presented the tool during AWS re:Invent 2023. A how-to on using an image generator seems redundant to us at this point. That is because Amazon is introducing its image generator late compared to cloud competitors Google Cloud, with Imagen, and Microsoft Azure, with Bing Image Creator. It is notable, however, that even after this long time, it is only possible to use the tool in English.
AWS has put this time into developing other features of Titan Image Generator. For example, users can use the tool only to adjust the background of a generated image. In addition, images from the tool will always have an invisible watermark. This makes it easier to check images on the internet for authenticity easily.
Also read: What can Amazon’s AI chatbot Q do?
In Amazon Bedrock, Claude 2.1 from Anthropic, Llama 2 from Meta, and the image generator Stable Diffusion XL 1.0 from Stability AI will also be available initially. As a result, the platform for finding AI models remains up-to-date with the latest developments.
More important is the introduction of the new Model Evaluation tool. Through this tool, users can more quickly find the foundation model that suits their needs. To do this, users choose a model from Bedrock and provide additional information about the usage goals for the model. To that, they add a dataset that will be used for the task. The tool will then automatically evaluate the available foundation models for accuracy, robustness and harmfulness. If these criteria are insufficient, a manual check can be done by an AWS team or by the requesting party itself.