If a picture is truly worth a thousand words, then Google Bard just got a substantial upgrade. Google’s generative AI chatbot will now supply pictures in response to your questions to make its answers clearer, and to “help you imagine more possibilities”.
Jack Krawcyzk, senior director of product at Google, gave a few examples of how it worked in a thread on Twitter.
📸 Bard (https://t.co/FycdN2l1HQ) starts its image journey today 📸
— Jack Krawczyk (@JackK) May 23, 2023
The first of many features we discussed at I/O is live right now: images in responses.
Enjoy some extra visual flair to help you imagine more possibilities.
Click an image to see the source for each one. 1/ pic.twitter.com/k4ruuhcELl
Bard will do this unprompted when appropriate, surfacing images from Google Image Search to accompany the copy it generates. For example, asking Bard to plan a sightseeing trip in London might include pictures of the must-see sights.
It generally works well, although Bard is still prone to make stuff up from time to time.
When The Standard asked Bard to name the best Derby County players of the 1990s, the bot included a photo of Mark Wright, while describing the player as a “key member of the Derby team that finished third in the Premier League in 1996-97”. Wright actually left the club in 1991, and Derby finished 12th in the season in question.
Still, if something doesn’t feel right, clicking on an image will allow you to see the source for each picture to dig deeper. This is especially handy if you’re asking for recipe ideas, as you’ll be able to find out more about whatever delectable-looking food Bard has surfaced.
You can also ask Bard for pictures directly. In a follow-up tweet, Krawczyk included an example of this, asking for vegetarian meals high in protein and fibre with “pics for added inspiration”.
Bard includes images in relevant responses and also when you specifically ask for them, starting with English.
— Jack Krawczyk (@JackK) May 23, 2023
In the coming weeks, Bard will be more visual:
We'll add support for more languages, ability to generate images, and prompt Bard with images using Google Lens.
2/ pic.twitter.com/tywKpQGA58
The update was trailed at Google’s I/O developer conference and more visual features announced at the event will be coming soon.
Specifically, Bard will soon be able to generate images in a similar fashion to the likes of Midjouney — and has already laid the groundwork to help users spot AI-generated fakes — and will let you prompt Bard with pictures of your own. On the latter, at I/O, Google showed one such use — getting Bard to write “funny captions” about a pair of dogs (your definition of “funny” will probably vary).
Bard’s image-sharing update is now live for English language responses wherever Bard is available. This notably doesn’t yet include the European Union, possibly for regulatory reasons. As Wired pointed out, at the time of writing more penguins had access to Bard than Europeans.