There are reports that Imagen 3 is letting users create images that resemble copyrighted characters, but this issue is far less severe than Grok’s lack of guardrails.
Google has shared the latest version of its Imagen 3 text-to-image generator across the US, letting more users test out its capabilities.
Venturebeat first reported that Google had quietly rolled out its latest version of Imagen through its ImageFX platform. Google DeepMind also shared a research paper on the AI model and said it is the company’s “best diffusion model for text-to-image generation”.
DeepMind said Imagen 3 is able to generate images with “better detail, richer lighting and fewer distracting artefacts” than its predecessors. Google first released Imagen in May 2022 to compete with OpenAI’s image creator – Dall E – and has been upgrading its capabilities as the market competition ramps up.
Users have been sharing their experiences with Imagen 3 online, with one user praising the quality but claiming it has more errors than the Imagen 2 model. Other users criticised the level of censorship built into the model’s guardrails.
The criticisms against Imagen 3’s guardrails come the same week that xAI released a text-to-image generator upgrade for Grok. This model appears to have very few guardrails, as images were quickly shared of copyrighted characters and public figures holding assault rifles or committing murder.
But Imagen 3 does not appear to have perfect guardrails when it comes to copyright protection. A report by The Verge claims the guardrails seem flexible, as it can be quite easy to generate images that resemble characters such as Sonic the Hedgehog, Mario and Mickey Mouse.
Meanwhile, advanced text-to-image generators have caused concerns around the growing sophistication of deepfakes. With the rise of this technology, experts believe AI could be used to influence elections and it is unclear if governments or tech giants can deal with the threat.
Earlier this month, a group of US senators introduced a No Fakes Act to make the creation of voice and visual likenesses of people, such as AI deepfakes, illegal without their consent.
Find out how emerging tech trends are transforming tomorrow with our new podcast, Future Human: The Series. Listen now on Spotify, on Apple or wherever you get your podcasts.