Products You May Like
Google has announced the rollout of a new version of its Imagen 3 AI image creator in Gemini. In addition to the usual spate of refinements and promises of prettier pictures, Google is once again allowing its model to generate images of people. This comes after it paused human generation in Imagen 3 in February after its initial release overemphasized diversity.
With the new Imagen 3 launch, Google promises better, more accurate depictions of your text prompts. The AI can again produce images of people, but there are still plenty of limitations there. The model will not make photorealistic images of identifiable people like celebrities and politicians. Likewise, it will block images that are “excessively gory, violent, or sexual.”
The Imagen model is generative AI, similar to Dall-E 3 from OpenAI or Stable Diffusion. After being fed a boatload of training data, the model can generate an image output from a text prompt. Shortly after its launch, users began noticing that Google’s attempts to filter out stereotypes may have gone too far. Prompts that should have produced men with generally lighter skin tones instead produced images of people (sometimes women) with darker skin.
One of the more popular examples of Google’s issues was asking the model to draw a US Senator in the 1800s—it now reliably draws old white guys instead of women and people of color.
Credit: Google
This screw-up became a cause célèbre in certain corners of the internet—Elon Musk, who is building a largely unrestricted AI for X (formerly Twitter), accused Google of being “anti-civilizational,” whatever that’s supposed to mean. Google was forced to backtrack due to the complaints, announcing in February that it would pause the generation of humans with Imagen while it retooled. Since then, attempting to have Gemini produce an image featuring people threw an error message.
Google says the updated model will roll out first to paying Gemini users, which will include those with Gemini Advanced, Business, or Enterprise accounts. However, it’s not live yet as of this posting. You can, however, try the new model in Google’s AI Labs. In our limited testing, the new Imagen is much less afraid to generate images of white guys. Google says it is still following its core design principles, and the images are watermarked with its SynthID system to help identify AI images.
The proliferation of AI image tools like Imagen 3 has led to widespread hand-wringing online as many predict an explosion of disinformation. Indeed, that may be in our future, but these tools are currently unable to produce convincing fakes—even the best AI images still look like they were constructed by AI. The tools are getting better quickly, as evidenced by Google’s rapid revamp of Imagen 3. So, that dystopian future may arrive sooner than we expect.