Google Makes a Host of AI Announcements

In its I/O event, Google talked about everything from updates to Gemini to an AI video generator.

Google’s I/O Developer’s Conference kicked off a week ago. Since then, the company has announced major updates for all of its products. The company’s AI ecosystem and product portfolio is proliferating at a breakneck speed. Here are the key takeaways summarized for you:

First up, they’re supersizing Gemini 1.5 Pro’s context window, doubling it to a whopping 2 million tokens. Large documents, tons of videos, a barrage of songs, entire codebases, you name it. You can now feed it all to Gemini and it will retain the information accurately.

Google also unveiled Gemini 1.5 Flash, a speed demon model built for efficiency, perfect for those who need answers fast. And the open-source enthusiast should get ready for Gemma 2, the next-gen model that’s about to hit the scene, along with PaliGemma, a fancy new vision-language model.

If you’re a Gemini Advanced subscriber, soon you’ll be able to create your own personalized AI buddies called “Gems,” just like ChatGPT’s GPTs.

Google also showed off Imagen 3, a text-to-image model that’s smarter and more creative than its predecessor.

Google also has a nice surprise for video creators. Its new Veo model (link) can whip up stunning 1080p videos over 60 seconds long, all from your text, image, or video prompts. You can try out a sneak peek of the tool on the Google VideoFX page, a tool that lets you create videos scene-by-scene and even add some sweet tunes to your masterpiece. VideoFX is in private preview for now, but ImageFX, powered by Imagen 3, is open for you to try.

But the real showstopper was Project Astra, Google’s prototype AI agent that can see, hear, and take action for you in real-time. Imagine a voice assistant that not only understands what you’re saying but also what it sees on your screen, whether it’s code, images, or videos. It can even reason and recall information like a champ. Keep your eyes peeled for Astra’s public debut in the Gemini app later this year. Chained actions can be taken by Astra. Though it might be scary for some, especially seeing how much power over our credentials, login info, and data we’re willing to happily surrender to AI agents, it’s also a breakthrough in its own right.

Google is also leveling up its “AI teammates” in Workspace, giving you virtual assistants that can tackle questions about emails, meetings, and more. And soon, you’ll be able to chat with Gemini in real-time, thanks to the upcoming Live feature.

Even Google Search is getting a makeover, with expanded AI Overviews, better planning tools, and AI-powered organization of search results. Gemini will even be able to plan and update your trip itineraries, while Search gets a “multi-step reasoning” upgrade to help you find answers faster. You can even ask questions about videos, and Search will analyze them to give you helpful AI Overviews. At the heart of Google’s core business of Search is this philosophy to satisfy a user’s query loop. That’s to say that the user should find the most relevant information or solution with one search. So far, the idea was to improve Google Search so much that the best page(s) gets shown at the top. For each query, there will be a page that solves it 100% without needing to re-search. That was SEO’s realm, with webmasters trying to game the system. It’s a whole thing. But now, the focus is shifting to AI.

By Abhimanyu

Unwrapping the fast-evolving AI popular culture.