As part of the Google I/O conference currently taking place in Mountain View, a new version of the generative neural network from the Imagen family was announced. The Imagen 3 algorithm is the most advanced image generator from Google to date.
Enhanced Capabilities and Features
The head of Google’s DeepMind AI research unit, Demis Hassabis, stated during the presentation that Imagen 3 understands test queries more accurately than the Imagen 2 model. He noted that the algorithm works more creatively and in greater detail, while also making fewer mistakes and creating fewer distracting artifacts.
To address concerns about the misuse of Imagen 3 for creating deepfakes, Google will utilize SynthID technology. This technology applies invisible cryptographic watermarks to media files, making attempts to use Google’s AI generator to create fake content effectively useless, notes NIX Solutions.
Accessibility and Security
Individual users can subscribe to Imagen 3 through Google’s ImageFX service. Developers and enterprise customers will have access to the AI generator via the Vertex AI machine learning platform. As in the past, Google did not share details regarding the data used to train the new algorithm. Rest assured, we’ll keep you updated with any new developments.