Maybe Google should call its annual developers' conference Google A/I instead of Google I/O, because the company is all about embracing generative AI this year. Google debuted its latest language model, Pathways Language Model 2 or PaLM 2, which now fully-powers Bard, Google's AI chatbot rival to Microsoft's ChatGPT-powered Bing chatbot.
Hsiao also said Bard will become "more visual" in the next few weeks, both in responses and prompts. This means Bard search results will start returning images for relevant queries, and also that people will be able to prompt bard "with images" using Google Lens integration. Hsiao demonstrated this with a photo of two dogs and the prompt "write a funny caption about these two," so you'll still need to prompt Bard with words in addition to the images (I'm not sure what I was thinking, but prompting Bard "with images" initially sounded much more exciting in my head).
Google will also be collaborating with third-parties to add features to Bard, such as text-prompted image generation thanks to integration with Adobe Firefly. The company is calling these "tools" and showed a graphic with several companies' logos including Kayak, OpenTable, Tripadvisor, Spotify, and Walmart. Finally, Google has dropped the waitlist and Bard is now available in 180 countries and has added support for prompts in Japanese and Korean.
Duet AI for Google Workspace
Google also demonstrated its new AI-powered toolset for Google Workspace (Google's suite of productivity apps, which includes Gmail, Docs, Slides, Sheets, and Meet), which it's now calling "Duet AI for Google Workspace" (the toolset, not Workspace). It's called "Duet AI" because Google imagines that you'll be collaborating with AI, like you collaborate with coworkers and other humans in Workspace apps right now.
Google actually first announced the upcoming Duet AI tools in a blog post back in March, but most of them — save for the writing assistant tool "Help me write" — are still unavailable to the public. The upcoming tools will let users do things like generate summaries from notes, create images from text prompts within Google Slides, create templates from text prompts in Google Sheets, and create custom backgrounds in Google Meet.
Google also demonstrated a future 'Sidekick' panel that will be able to read your content and "engage" with you using contextual prompts. Google Workspace VP and GM Aparna Pappu demonstrated this feature with an unfinished children's story. Asking the Sidekick for help (don't worry, it's not going to prompt you without solicitation) generated some questions about what might happen next.
Most of the new Workspace tools aren't available to the public yet, and will be rolling out selectively over the next few months. If you're interested in potentially getting early access to test out these features, you can sign up for the Google Labs waitlist here.