Gemini’s Image Generation Capabilities Not As Strong As They Used To Be Thanks To A Nerf From Google

Feb 24, 2024 at 04:22am EST
Galaxy S24 and Pixel 8 Series Could Soon Get On-Device Web Page Summarizing Feature Thanks to Gemini Nano

Gemini, Google's generative AI chatbot, recently went through a debacle where it ended up generating inaccurate and offensive imagery. Following the instance, the company has decided to nerf its image generation capabilities until a fix can be found and services can be resumed.

Gemini lands itself in hot waters over some troubled image generation of people of certain ethnicities

Google talks about how Gemini is a separate product entity, and it is not directly linked to Search, its AI models, and other services that it provides. Once the app was ready, Google manually tuned the model so the image generation could be diverse instead of showing bias towards certain groups of people or ethnicities. It makes sense for Google to do this to ensure that users from different areas of the world can use the app without running into any major issues.

Related Story Apple Removes The Fog Around Its New Cloud-Based, And 20-Billion-Parameter On-Device AI Models, Brushes Aside Google’s Contributions While Hyping NVIDIA’s

However, none of the countermeasures could have prevented what happened afterward. Gemini could not account for use cases where the users were actually looking for a certain ethnicity to be present in an image. This led to some really strange prompts, such as a "Black teacher in a classroom," which caused the model to generate incorrect as well as offensive images. In some cases, it even refused to generate images to begin with. The reason is that the model considered them sensitive even though they weren't.

So, what is Google's stance on this whole situation? Well, the company has gone ahead and shut down the ability to generate images of people in Gemini. The company has talked about how it will enable it once the issue is fixed after the whole model goes through extensive testing. Google has also talked about how AI models are not perfect, and despite being advanced from when they were introduced, they still have the tendency to hallucinate, which can cause issues.

As for the timeline, Google has not shared one, and we will have to wait for the Gemini image generation to be up and ready. We do hope that when it does come back, it comes back better than before and improved at creating images.

News Source: Google Blog

About the author: I have been tinkering with Android devices ever since the early days of the HTC Desire. Over time, I have grown a fondness for the ecosystem and now I cannot live without it. Although some might believe that I have sold my soul to Android, but I believe it is not the case. You can find me writing tutorials and posting guides on a number of different smartphones. When I am not writing here, I am wasting myself away in books, journals, or on Steam.

Follow Wccftech on Google to get more of our news coverage in your feeds.

Deal of the Day