Last December, we had to visit specific websites in order to have our natural language prompts turned into created AI art, but not anymore! Google revealed on Thursday that users who have chosen to participate in its Search Generative Experience (SGE) will have access to the ability to produce AI images right from the default Search bar.
SGE is Google’s idea for the future of internet search. The system will employ the same data that the list’s links lead to in order to synthesize a (reasonably) logical response to the user’s natural language query as opposed to selecting websites from a returned list. The revisions on Thursday are a logical extension of that experience, essentially returning created images instead of generated words (using the company’s Imagen text-to-picture AI). Users enter a description of what they are looking for (in Google’s example, a Capybara preparing breakfast), and the search engine generates four options for them to choose from and further refine. Additionally, users will be able to download or export the created photographs to Google Drive.
Additionally, users will have the option to create images right in Google Images. So, if you’re searching for (again using Google as an example) “minimalist halloween table settings” or “spooky dog house ideas,” you can type that into the search field and Google will produce a picture based on that. The coolest part is that you can use Google Lens to look for actual, real-world things that most closely reflect the hallucination the machine created for you.
Of course, the new features have a few restrictions in place. To utilize any of this, you must be a Google Labs member and have chosen to participate in the SGE program. The new picture production features will also only be accessible to users over the age of 18 in the US, in English-language applications, and within the country. Given that Google just went above and beyond to make the program available to teens, this additional condition is a little bit strange.
Additionally, the business is stepping up its efforts to prevent the abuse of generative AI technology. Users will be prevented from producing photoshopped pictures of people’s faces. There is no issue if you want a photorealistic capybara cooking bacon. If you want a photorealistic image of Colonel Sanders cooking bacon, you’ll run into problems that go beyond the bounds of accepted commercial practice. Additionally, you won’t be able to create photos of “notable” individuals, so Colonel Sanders is already out.
Finally, Google has begun rolling out the SynthID technology that DeepMind unveiled last month. In addition to identifying a generated image as such and providing information on who, or what, developed it and when, SythID is a visually imperceptible metadata watermark. Select users will be able to use the new features today through the Labs page (click the flask icon) in the Google app for iOS and Android, as well as on Chrome desktop. More users will be able to access the features in the upcoming weeks.
