Google is expanding its AI capabilities with the release of Imagen 2, a new and improved version of its image-generating model. This innovative technology, which allows for the creation and editing of images based on a text prompt, will now be more widely available to those using Vertex AI through Google Cloud. However, while this may be exciting news for some, there are several important details that Google is not disclosing.
One of the main concerns is the source of data used to train Imagen 2. Google is choosing not to reveal this information, raising questions about potential contributors who may not be aware that their data was used or have no way to opt out or receive compensation for its use.
“Called Imagen 2, Google’s enhanced model – which was quietly launched in preview at the tech giant’s I/O conference in May – was developed using technology from Google DeepMind, Google’s flagship AI lab.
Compared to the first-gen Imagen, it’s “significantly” improved in terms of image quality, Google claims (the company bizarrely refused to share image samples prior to this morning), and introduces new capabilities including the ability to render text and logos.”
Along with the ability to add text and logos, Imagen 2 can also generate images in multiple languages and overlay logos onto existing images, setting it apart from other leading image-generating models like OpenAI’s DALL-E 3 and Amazon’s Titan Image Generator.
- Chinese
- Hindi
- Japanese
- Korean
- Portuguese
- English
- Spanish
In addition, Imagen 2 has improved multilingual understanding, thanks to novel training and modeling techniques. This allows the model to translate text prompts from one language to an output in another language.
Another notable feature is Imagen 2’s use of SynthID, a method developed by Google DeepMind to apply invisible watermarks to the images it creates. While these watermarks are resilient to common image edits, detecting them requires a tool provided by Google, further limiting access to the full capabilities of Imagen 2.
“But as policymakers express concern over the growing volume of AI-generated disinformation on the web, it’ll perhaps allay some fears.”
However, with the ongoing legal debate around whether AI vendors can legally use publicly available data for commercial purposes, Google is taking the safe route by not disclosing the data used to train Imagen 2. This is in contrast to its approach with the first-generation Imagen, where it revealed that a version of the LAION data set was used, containing problematic content such as copyrighted artwork and altered celebrity photos.
While some AI companies, like Stability AI and OpenAI, allow creators to opt out of their training data sets and others like Adobe and Getty Images offer compensation schemes, Google does not currently offer any mechanism for creators to opt out or receive compensation.
- Original data not disclosed
- Debates around fair use
- Lawsuits in progress
- No opt-out mechanism for creators
Instead, Google provides an indemnification policy for eligible Vertex AI customers, protecting them from copyright claims related to the use of training data and outputs from Imagen 2. This is a concern for companies and developers, as studies have shown that AI models are not immune to regurgitation, meaning that they can spit out almost identical copies of training examples.
“An academic study showed that the first-gen Imagen wasn’t immune to this phenomenon, spitting out identifiable photos of real people, copyrighted work by artists and more when prompted in specific ways.”
In fact, a recent survey found that nearly a third of Fortune 500 companies listed intellectual property as their top concern regarding the use of generative AI, showcasing the importance of protecting original content. It is a concern that Google hopes its expanded indemnification policy will help to address, as it now covers Imagen 2 outputs as well.
However, for creators who may have unintentionally contributed to the data set used to train Imagen 2, there is currently no recourse for opting out or receiving compensation. As Google continues to advance its AI technology, it will be interesting to see if and how it addresses these concerns from creators and policymakers alike.
