Google I/O 2017 Keynote – AI, Machine Learning and more

AI first and everywhere

Right from the start it became clear that Machine Learning and AI were to become the main theme of the Google I/O and should be seen as the driving engines for future advancements and technological progress.

Google CEO Sundar Pichai set the stage for the keynote presentations. Just like last year he pointed at the importance of speech and image recognition. These are the most natural modalities and had only recently become widely available as interfaces to digital products. It is pretty obvious that especially these modalities must be backed by powerful AI and hence Machine Learning capabilities. He summed it all up with one simple but nevertheless intriguing phrase:

Mobile first to AI first. And after that almost every other speaker emphasized the fact that all the Google services are, in some way or another, powered by AI and Machine Learning at a large scale. So it’s not only AI first but also AI everywhere .

Announcements

So, that said, here is my little overview of some of the thing I found most interesting.

AI and Machine Learning

Since we are AI first, here come the announcements around AI and Machine Learning:

AI first data centers: data centers build from ground up with AI and Machine Learning in mind

Google Cloud Machine Learning Services, launched a little earlier this year, now enables us to harness the power of Machine Learning and Deep Learning in the Google Cloud (last year we begged for it)

TensorProcessingUnits coming to the Cloud Compute Engine

TensorFlow Lite on Android: processing directly on the device

Google.Ai: new platform for information about AI and Machine Learning

Side note: they have built neural networks which are trained to build new neural networks to be trained with actual data (hope its not called Skynet)

Other Services and Features

And finally here is the best of the other announcements not revolving around AI or Machine Learning:

Google Lens: new service with powerful location/context sensitive image recognition used standalone or in Google Assistant

Google Home: finally coming to Germany (YEIH!), new features like free calls (US/Canada only ), visual response to phone or Chromecast and recognizing persons by voice and using this information appropriately

Google Assistant: coming to the iPhone

Google Assistant SDK: enables us to embed Google Assistant features on custom devices, i.e. RaspberryPi 3

Actions on Google: finally we can add our own services to Google Assistant or Google Home

Fluid Experiences and Vitals for Android O: many improvements regarding usability, security, stability, performance, battery consumption

Android Go: a light version of Android for low budget phones with less powerful hardware (i.e. RAM, CPU) targeting emerging markets like Brazil and India

Android Ecosystem: first-class support Kotlin, tons of improvements to the AndroidStudio, better profiler, Vitals within the Google Play Developer Console to better understand real problems users of the app are experiencing

Visual Positioning System: indoor positioning using visual features of the room, this is huge for VR and AR alike

Standalone Daydream VR headsets: no extra phone, no extra computer no cables

Wrap up

There was much, much more and this is just what struck me as the most interesting. As well as the paradigm shift from Mobile First to AI First which still resonates with me.

I am really looking forward to get my hands dirty, especially with Machine Learning services the Google Assistant SDK and Actions on Google.

I’ll keep you posted.