Amazon has announced the release of an image generator, joining a host of tech giants and startups that have already launched their own versions.
During a keynote at its re:Invent conference, Amazon unveiled Titan Image Generator, which is now available in preview for AWS customers on Bedrock, Amazon’s AI development platform. As part of Amazon’s Titan family of generative AI models, Titan Image Generator can generate new images based on a given text description or modify existing images.
VP for data and machine learning services at AWS, Swami Sivasubramanian, highlighted the model’s capabilities, stating, “You can use the model to easily swap out an existing image background to a background of a rainforest, for example. You can use the model to seamlessly swap out backgrounds to generate lifestyle images, all while retaining the main subject of the image and to create a few more options.”
According to Amazon, Titan Image Generator was trained on a diverse set of datasets across a broad range of domains and can be fine-tuned on custom datasets. The model includes built-in mitigations for toxicity and bias, although the effectiveness of these mitigations is yet to be determined through testing. Amazon did not disclose the sources of the training datasets or whether permission was obtained from or compensation provided to the creators of the images used to train Titan Image Generator.
Some companies developing image generators, such as Stability AI and OpenAI, allow creators to opt out of training datasets. Others, like Adobe and Getty Images, are implementing compensation schemes for creators, although the details of these schemes may not always be well-compensated or transparent.
Sivasubramanian assured that Amazon will protect customers accused of violating copyright with images generated by Titan Image Generator, in line with its AI indemnification policy, providing reassurance to AWS customers concerned about unintentional duplication.
In a recent survey of Fortune 500 companies by Acrolinx, nearly a third expressed that intellectual property was their primary concern regarding the use of generative AI. Another poll found that nine out of 10 developers “heavily consider” IP protection when deciding whether to use generative AI.
Sivasubramanian emphasized that Amazon is carefully selecting how it trains its models and the data used for training. Furthermore, images created with Titan Image Generator will feature a “tamper-resistant” invisible watermark by default to prevent the spread of AI-generated misinformation and abusive imagery.
Watermarks are part of Amazon’s voluntary commitment around AI, which it signed with the White House, and other signatories like Google and Microsoft have also adopted or developed their own solutions to address the problem of content provenance and authenticity.