Google continues to push the accelerator of artificial intelligence: in the past few hours it has announced several new features coming and being distributed for one of its most popular software ever, Maps, and for its experimental virtual assistant Bard, which now “knows how to draw”. But let’s go in order.
The Mountain View giant intends to implement generative artificial intelligence that helps the user discover new places according to their needs and tastes. In a nutshell, it will be an LLM (Large Language Model), i.e. an advanced chatbot capable of understanding user requests and providing relevant answers based on a wide range of parameters.
The algorithm, in Google’s words, will analyze “Maps’ insights into more than 250 million places and trusted information from our community of more than 300 million contributors to quickly provide suggestions on where to go.” It will initially be available in the U.S., and there’s no timeline information for other countries.
Google says it will proceed with great caution in rolling out the feature: first it will only make it available to Local Guides, power users, part of the Maps community and therefore more likely to have a collaborative approach. For the time being, there is no more detailed information about how the interface will work – and especially how much better the user experience will be compared to the currently available search system.
Google’s Gen AI chatbot finally goes “multimodal” – one of the latest “buzzwords” to emerge with this technology, which simply means that the agent is able to generate different types of content. Bard has been dealing with text so far, and now it can also create images.
Multimodality is one of the fields where the various Gen AI agent developers compete with each other a lot; in this sense, Google is a bit of a catch-up, since its main rival ChatGPT has been able to generate text and images for a few months now (but for a fee, where Bard is free). Google explains that Bard relies on the Imagen 2 algorithm. Of course, you always proceed at a prompt: just write to the chatbot what kind of image you want to generate, specify parameters, styles and so on, and wait a few seconds for the result.
Google explains that it has implemented the feature with an eye to security and safeguarding user privacy: watermarks that confirm the AI origin of the image are inserted directly into the pixels that compose it, and there is a whole series of protections against abuse of the function to generate images of famous people and limit violent content, offensive or sexually explicit (a clear reference to the Taylor Swift case that exploded in recent days).
That’s not all: a few months ago Bard was updated with a more advanced AI algorithm, Gemini Pro, but it was only compatible with the English language. Gemini Pro now supports 40 different languages, and is available in over 230 markets up from 170. However, image generation only works in English for now. By the way, Google has also released an image generation tool based on the same Imagen 2 algorithm in its own right: it is called ImageFX and joins the pre-existing MusicFX and TextFX. You can test them on Google Labs. Finally, Bard now supports source verification in other languages as well.