At its latest I/O 2025 event, Google made a big proclamation: AI is no longer something that simply runs on software, it’s fundamentally changing the way we communicate, co-operate and interface with the digital world.
The theme of Google’s new lineup of products is the fusion of the company’s advanced artificial intelligence with the real world, from more realistic virtual video calls to almost instant translation to, yes, an even more cloud-attuned virtual assistant. Here are the most notable AI announcements from the event.
The spotlight was on three major updates
- Beam – From 2D Video to 3D Reality
Google and HP launched Beam – a 2D to 3D video communication platform of the future. It’s video calling, but with the real-life 3D presence.
Beam captures and displays people in real-time using AI, purpose-built cameras, and a 3D light field display to offer a more human and more immersive conversation. Early users are supposed to get it later this year - Translations of what people say on the fly in Google Meet
Google Meet is about to get a powerful upgrade. AI Pro and Ultra services users can now listen to almost real-time English to Spanish translations. But this is no ordinary robotic text-to-speech.
The system will also mirror the speaker’s intonation, emotions and even gestures, to make translated conversations sound that much more human. Workspace business customers will also be given early access for testing, with additional languages to follow - Project Astra Arrives at Gemini Live
Project Astra, Google’s supercharged AI assistant with the ability to share its camera and display capabilities is now available in Gemini Live.
That lets people, when the camera is on, interact with the world by asking questions about what they see or get assistance when sharing their screen. It’s already rolled out for Android users, and iOS users will see that support over the next few weeks.
Why does this matter?
First Google isn’t only a new supplier of AI tools it’s redefining how we fundamentally interact with information and each other.
- Beam could revolutionize the way we communicate remotely, delivering a level of presence not seen before.
- Google Meet translations will knock down language barriers, particularly among global teams and in international communication.
- Gemini Live with Astra is Google’s approach to real-time smart assistance which is aware of context and deeply integrated into user workflows.
And these tools aren’t only for flexing: They are practical upgrades that could soon be necessary for how we work, learn, and connect with one another.
Other key announcements at I/O 2025
- Gemini 2.5Pro is top of the LMArena leaderboard, having benefited from major work on performance and AI.
- Google’s latest generation TPU (Tensor Processing Unit) Ironwood provides not only a 10x boost in performance, compared to the previous generation, but it also delivers 42.5 exaflops per pod for the next wave of inferential AI.
- Chrome, Search and the Gemini app will soon introduce agentic AI capabilities, for more personalized and autonomous browsing and search experiences.
With these updates, Google isn’t just playing catch-up in the A.I. arms race. it’s taking drastic measures in the right direction. From Beam’s 3D immersive calls to Gemini’s boost in intelligence and Google Meet’s lifelike translations, the message is clear: AI is no longer in the lab, it’s in your daily life.
- Also Read: Sundar Pichai at Google I/O 2025: “World Adopting AI Faster Than Ever” as Google Glasses Return
And with Pichai playing up getting model improvements out the door quickly and infrastructure advancements, Google looks to be committed to shipping enhancements all the time, not just in one big annual event.
Quick Recap
- Beam transforms any 2D video into a 3D video with a remarkable combination of AI and HP.
- Pro and Ultra users get natural real-time translations in Google Meet beta.
- Gemini Live’s screen grabs and smart camera capabilities receive a boost from Project Astra integration into Gemini Live.
- Gemini 2.5 Pro leads all AI benchmarks this month, with an Elo increase of 300+.
- Ironwood TPU v7 AI performance per pod peaks at observatory levels of 42.5 exaflops per pod.
