AI tools are here to stay, helping us search the internet, decide what to wear, enhance visual effects in movies, find better jobs, and more. As time goes by, these tools will become smarter and start to perform additional functions – such as the ability to scan the web for image searches.
From today, it’s not just textual results. Just as we search for images on Google, now we can search for images using Google’s AI tool, and it will display the most relevant images within the chat based on our requests.
Google updated this week that images can ‘bring concepts to life, make recommendations more convincing, and enhance responses when you request visual information.’ In other words, it can add a bit more flair to the text produced by Bard, whether it’s explaining scientific concepts or recommending restaurants.
At the time of writing, generating images is not something ChatGPT can do – OpenAI’s bot is entirely text-based, although it can present text in various formats, including tables.
Currently, there is no toggle switch for enabling or disabling images in the regular results produced by Bard when you ask it a question: the bot will simply include images if it thinks they will be helpful.
Some of the searches we ran that came with accompanying images were for recommendations for bars, an explanation of how optical illusions work, and a query about the first working airplane – questions that require visual reinforcement to explain the answer.
When we asked for an explanation of what DNA is, or a summary of the Battle of Hastings, images were not included in Bard’s output. However, from now on, you can specifically request an illustration or relevant image in your query, and Google Bard will provide suitable images.
This also applies to answers that do not have a clear image to show – if you are looking for information on theories or philosophies, for example. Google Bard will try to bring up something relevant, even if it doesn’t quite hit the mark.
All images you see in Bard will come with captions and credits attached, and if you click on the credit, you will be taken to the internet page where the image source is located.
Bard does not limit itself to images without copyrights or images published under a Creative Commons license. If an image is found on the internet, Bard will use it. It’s also worth noting that when you scroll through the multiple drafts Bard offers, the images tend not to change, even when there are variations in the text.
Like on any webpage, you can right-click on the images in Bard’s responses to save them. However, it’s probably a better idea to click on the website where the image is from, which usually provides a higher-resolution version.
In addition to embedding images in its results, Bard can also directly search the internet for images – just like a Google image search. Bard simply finds images, not generates them like tools such as DALL-E.
So, for example, you could ask Bard to show you pictures of white roses, or old dogs, or a sunset on the beach. You can specify the number of images you want.
The images are generally of good quality and well-suited to the instructions, based on our tests. You can be quite specific – for example, ‘Tom Hanks looks confused’ or ‘a white door on a red wall’ – and Bard will do its best. However, Bard is limited by the captions and tags of the original images.
When searching for specific images, carefully read the description written next to each one, it will give you a bit more information about what you are looking at: it could be a description of what is in the image, or it may tell you about the specific type of flower, tree, or animal.
**Stay updated on everything new in the world of AI>> Join our community, leave your details, and we will update you directly via email.**
