Google Bard may have only just launched, but Microsoft has clawed attention back to its own Bing with ChatGPT AI with the imminent release of a long-awaited feature: AI image generation.
AI art generators have been popular for a while now but what makes Bing with ChatGPT different is the potential for users to create changes through a conversational thread. The same as you might ask a human artist to add a few tweaks here or there.
For example, Corporate Vice President at Bing and Microsoft Jordi Ribas demonstrated the feature by getting the AI to draw a picture of a Shih-Tzu sailing a boat.
The AI completed this as expected but then Ribas asked the AI to “add university of Michigan colors” which then dressed the dog in purple and yellow apparel and recolored the boat. A typical AI art tool wouldn't know what the colors of the university were and would require them to be explicitly explained.
Excited to share that Bing Image Creator is rolling out to the new Bing and Edge preview today. Now you can create an image by simply using your own words to describe the picture you want to see. (1/4) pic.twitter.com/kSmvopFJOFMarch 21, 2023
In a blog post Microsoft also demonstrated the AI could switch between image and text responses in the same conversation. While text responses are powered by GPT4, image creation is handled by OpenAI’s own image tool DALL∙E.
While we couldn’t get Bing chat to create images just yet, you can try a preview of its image-generating technology now. Interestingly one New York based startup has created an AI that can generate videos from text.
Bing vs Bard
With Google Bard now starting to be released to the public (you will still have a join a waiting list), we can start making more direct comparisons between the two.
Bard currently has no multimodal capabilities like image generation but it doesn’t have a turn limit. This suggests that Google has more faith in its chat model than Microsoft who limit Bing conversations after it found longer chats could cause it to go off the deep end.
In terms of text responses, both chatbots seem pretty confident but actually use two different models. Bing uses GPT4 while Bard employs Google's own LLAMDA model.
However, until Bard can match Bing’s imaging efforts, it won’t be as useful (or as fun) for creatives.