Google is working on a countless number of projects, and I/O is, as always, an exciting time to see what’s happening at the tech giant. This year’s Google I/O keynote was much like their previous keynotes - lots of projects with a wide variety of impact.
While critics could complain that there was no big release during this keynote, not even Android P, I think the scattered approach to bring ML to literally everything Google makes is a big enough deal in itself. Google has been crunching data at godspeed with its custom Tensor Processing Units (TPUs). Today, they launched the v3 of their physical TPUs, which have some fancy liquid cooling for significant gains in performance. Google’s approach of adding AI to almost all applications, including Google Photos and Maps, is surely going to pay off - with millions of people already depending on it, it will save time and money.
All that said, my favourite part of this event is the new Google Assistant appointment booking feature.
When you ask Google Assistant to set an appointment with a business that doesn’t support online reservations, the Assistant will call the business and speak to a person in a real conversation. It sounds sci-fi, but the Assistant can actually converse beautifully. Take a look:
(The video below starts playing at 1:55:40 seconds)
The assistant handles this conversation just like a human. But what’s more impressive is the way it handles non-ideal situations:
(starts playing at 1:57:56)
Now, Pichai did confirm that this does not always work. Nonetheless, the potential is huge. Using TensorFlow to train a speech synthesizer - something that Google’s DeepMind has spent years on, WaveNet - with millions of phone calls is already a great start.
Google I/O had a few more interesting bits I would like to share. The company is also bringing AR to maps, to make directions better and having a “visual positioning system” VPS. They are also bringing huge improvements to Google Lens. See this really cool demo here:
(starts playing at 2:45:26)
And finally, on Android P, they are pushing hard on Digital Wellbeing - managing screen time and app usage, while handling annoying notifications better.
(starts playing at 2:31:05)
All in all, Google’s keynote this year was a fantastic peek into what’s coming in our lives. As a computer science student, it inspires me to deep-dive into machine learning and make something cool with frameworks like TensorFlow.