Google’s Enhanced Image-Generating AI: What’s New and Improved

Google Enhances Image Generation Technology to Compete with Rivals

At the recent I/O developer conference in Mountain View, Google unveiled Imagen 3, the latest iteration of its Imagen generative AI model family, reinforcing its commitment to staying ahead in the competitive AI landscape.

Demis Hassabis, CEO of DeepMind, Google’s AI research division, emphasized that Imagen 3 offers a significant improvement in interpreting text prompts compared to its predecessor, Imagen 2. He noted that this new model boasts enhanced creativity and detail in its image generation, while also minimizing “distracting artifacts” and errors. Hassabis stated, “This is our best model yet for rendering text, an ongoing challenge for image-generation technologies.”

To address concerns regarding the potential for creating deepfakes, Google has integrated SynthID, a solution developed by DeepMind that applies invisible, cryptographic watermarks to media, ensuring authenticity and accountability.

For those interested in experiencing Imagen 3, sign-ups for a private preview are available through Google’s ImageFX tool. The company has announced that this model will be accessible soon for developers and corporate clients utilizing Vertex AI, Google’s enterprise-grade generative AI platform.

While Google maintains a level of secrecy regarding the data sources used to train its AI models, it is known that much of the training data is sourced from public websites, repositories, and datasets across the web. However, this has raised questions due to the inclusion of copyrighted material scraped without permission from content creators, which has led to various intellectual property lawsuits.

Google provides web publishers with controls to prevent data scraping from their sites, including images and videos. However, unlike some competitors, Google has not introduced an "opt-out" tool or pledged to compensate rights holders for their potentially unknowing contributions to the training data.

While this lack of transparency is not unexpected, it is disappointing, especially coming from a company with resources as vast as Google’s.

Stay tuned for our upcoming AI newsletter! Sign up here to receive the latest updates directly in your inbox starting June 5.