The big picture: Major tech companies are quickly adopting generative AI as a new front in the productivity and search engine markets. Not long after Microsoft and Google unveiled their takes on AI chatbots, Adobe and Microsoft have revealed their plans for AI-based image generation.
Microsoft and Adobe unveiled their first applications for visually-based generative AI on Tuesday. The two companies have different intentions for the technology, but both are starting with a slow rollout in light of concerns over the ethics of generative AI.
The Redmond company announced Bing Image Creator, an image generator it’s integrating into the recently released Bing AI chatbot. Users can enter prompts into the chatbot window to create images through an enhanced version of the DALL-E AI model. However, DALL-E competitor Midjourney recently gained a leg up in realism with version 5, for now at least.
Bing Image Creator is currently only available in Bing preview’s creative mode but is eventually coming to balanced mode, precise mode, and Microsoft Edge. When it arrives in Edge, users can access it by clicking on the browser’s Copilot sidebar.
Meanwhile, Adobe revealed a family of generative AI models collectively named Firefly, with a beta available now. The AI integrates with the company’s image editing software to let users create pictures, and text effects using text prompts.
Adobe will roll its AI image generators into Creative Cloud, Document Cloud, Experience Cloud, and Adobe Express workflows. The first programs to receive it will be Adobe Express, Experience Manager, Photoshop, and Illustrator.
Concerns over the creation of harmful images and the use of copyrighted work to train AI models have made the technology as controversial as it is intriguing. Microsoft and Adobe are trying to address those concerns with measures to assure users they aren’t taking AI lightly.
Microsoft is implementing a staggered release for Bing Image Creator to see what its users do with the new tool. It will also include guardrails to ensure Bing doesn’t generate harmful pictures. It will also embed watermarks to identify images as being AI-generated.
Adobe is taking multiple steps to avoid copyright infringement and let creators opt out of helping to teach AI models. The company trained its first AI on Adobe stock photos, openly-licensed work, and public domain material. When Firefly exits beta, Adobe will unveil how it intends to compensate stock photo makers for their contributions to the AI.
Adobe will also refrain from using current Cloud subscribers’ content to feed its machine-learning models. Eventually, the company wants to convince its users to help train its AI but will let them assign a “Do Not Train” tag to their work. The creation will carry the anti-training tag anywhere it is stored, published, or otherwise used.
source : www.techspot.com