Microsoft’s AI-powered image generator wants to spark the imagination of its users. Dubbed Image Creator, the tool is online now, in preview, in select markets.
Microsoft has begun rolling out Image Creator for Bing as a preview in select markets, preparing the AI-powered art builder for a wider rollout in Edge later this month. In a related blog post and video, the company showed how Image Creator will work and further explained the limitations it will place on the prompts users generate. Several weeks ago, Meta already demonstrated its powers in this area with Make-A-Video, which tackled the creation of images from two different images, their animation, but also the creation of a GIF from a text prompt. A week later, it’s Google’s turn to animate text with Imagen Video.
Last week, Microsoft said it would introduce an AI-powered image generator for Bing and Edge using the more advanced DALL-E 2 algorithm. It appears that Image Creator will be available from Bing .com and that a related version will be available from Edge soon after. The editor showed that Image Creator works in Edge’s sidebar, cutting out a small vertical column to display search results and other information, as well as handy utilities. This is where everyone will be able to access the new image creator.
A limitless tool for the imagination
In a video, Microsoft showed how users could generate a prompt using conventional expressions such as art styles. Image Creator can thus return several small results in just a few seconds. It is not yet known if there will be some kind of credit system or some other counter to limit the generation of prompts for any user. Here, Microsoft’s approach is more collaborative: the example shown is a user designing a “dream house” using Image Creator’s content creation tools and then sharing it on social networks. Again the image appeared within seconds and four images were generated.
Microsoft’s blog post suggests that the AI imaging tools will work similarly to other services like Midjourney or DreamStudio that run on the Azure cloud. “We’ve found that Image Creator generally works best when you write a description of something, with additional context like the location or art style you want to emulate, as opposed to a more limited description,” Microsoft said.
Responsible AI highlighted
The firm will also use artificial intelligence to filter requests, using the same kinds of signals that help Microsoft Defender filter out problematic websites, for example. These blocklists and classifiers will be used to “reduce the risk of using offensive prompts”. Interestingly, Microsoft is also applying additional technology to address biases found in AI image generation. (He didn’t elaborate on what that means, although anecdotally some generic directions seem to favor results with certain skin tones).
“We take our commitment to responsible AI seriously,” Microsoft said. “To help prevent the delivery of inappropriate results across the Designer app and Image Creator, we are working with our partner OpenAI, who developed DALL∙E 2, to take the necessary steps and we will continue to develop our approach. We will regularly consider the feedback we have and share it with OpenAI to improve the model as well as apply it to our own remediation work.” This partnership could be strengthened at the capital level. According to The Information and the Wall Street Journal, OpenAI is in discussion with Microsoft about an additional investment (the Redmond company has already invested $1 billion). Finally, the publisher said its image generations would be governed by its content policy, which prevents images of child sexual abuse, intimate activity without consent, suicide, terrorism, hate speech, etc.