One of the most anticipated yearly tech events is Google I/O , an annual software developer-focused conference held by Google in San Francisco, California. Google I/O features highly technical, in-depth sessions focused on building web, mobile, and enterprise applications with Google and open web technologies. This year marks the 8th edition since the first one back in 2008.
In this year’s almost 2 hrs keynote, current trends , especially in mobile where 20% of Google searches on Android platforms are voice queries rather than text-input are covered . Google’s search engine capabilities are second to none & future developments will place more emphasis on embedding deep machine learning or artificial intelligence to predict what you want to search for (before you even know it?). A ‘knowledge graph’ can be built & individualized , connecting People to Locations to Things & developing relationships between them. With built-in AI in Google photos (click here to check it out) , you try to search for a place or object, such as “beach”. This will list out all your photos that taken on a beach! This also works on videos. You can now perform translation from text, voice & even visual (optical character recognition) through apps. With “image” or “voice” recognition technologies built-into apps, you can search for anything other than text ,accurately. Other apps or products introduced during the keynote includes,
- Google Assistant(previously Google Now) – Personal digital assistant (not unlike Siri on iOS or Cortana on Windows) this is already integrated into latest Android 6.0 or later versions. The demo during the keynote shows how to can talk naturally to Google Assistant to make a reservation through an app (without opening the particular app for it). Technology behind it(natural language processing) have been involving from chatbots that can simulate chats using AI. Here’s a video comparing all the 3 digital assistants.
- Google Home – the same technology used in assistant is integrated in an IoT device that has a built-in speaker, microphone & internet connectivity. With Home, you would not need to carry your smartphone around to interact with apps & services by voice. This is basically same the concept as Amazon Echo released in 2014. It will be interesting to see how this compares with Echo & also how Google addresses privacy concerns like eaves-dropping on sensitive conversations.
- Allo – A whatsapp/skype/messenger-like app but with “assistive” response, basically helping you to sound more intelligent with AI?
- Duo – Video calling similar to facetime in iOS. One interesting feature is that it will start streaming video before you respond to pick up the “call”. One common issue with video calling apps is the consistent quality/reliability streaming video (lag) & audio across different networks, hopefully Allo will have a robust streaming engine to deal with these issues.
- Android N – Android is by far the most popular operating systems for phones, media players, ICE etc. with Google Play store having 65B installations in the last year alone. Google usually announces new Android versions following dessert names (last 2 versions were called “Lollipop” & “Marshmallow”) . However, due to difficulty ion finding a dessert name beginning with “N”, they have left it to developers to name it. Recently, there was a spate of malicious apps appearing on Appstore & Play. With “SafetyNet” cloud based analytics , apps on Google Play will be monitored & “misbehaving” apps will be flagged. Other features covered include “split-screens multi-tasking” which enables you to look at multiple apps in the same screen (probably only practical on phablets due to the display real-estate, unicode-9 emoji standard support with politically correct emojis with different screen tones. You can read more about emojis here. Another feature is the update process which will be done in the background & without re-logging into your smartphone.
- Daydream VR – VR or virtual reality have been gaining a lot of interest lately with product launches like Occulus Rift , Sony VR, Samsung Gear VR etc. Google released Google Cardboard ,a virtual reality (VR) platform developed by Google for use with a smartphone previously, a very low-cost VR solution using cardboard cut-outs with lenses (or you can also purchase ready-made plastic head mounts). Due to differences in quality of sensors in smartphones, the experience can be barely usable in some cases using low-end SOC’s. Daydream bridges the development experience by streamlining the tools & development (including a VR-enabled Play store) , hardware compliance/requirements & libraries making it easier to develop Daydream ready apps.
- Google Wear 2.0 – latest smartwatch OS , new features are stand-alone apps that does not need to be connected to your smartphone to use. Mostly, you would need to wait for more power smartwatches with built-in cellular connectivity to take advantage of this.
- Android Studio 2.2 – Native development IDE for Android OS. New features includes improved emulator performances to test apps in development, improved layout designer, apk analyser.
- Instant Apps – Run Apps without installation through the Google Play store. Basically, this works like a browser-based application where it downloads partial functionality/modules of a particular app & running it without installing an app. This works on Android version since Jellybean.
In summary, the big theme was the application of deep machine learning & AI & also extension of Google platform to IoT devices.
Tensor Processing Unit or TPU’s are new processors develop by Google to handle the massive computing needs for general purpose AI. This is used in Alpha-Go,the first computer that mastered the game of “Go” , a game of profound complexity. There are more possible positions in Go than there are atoms in the universe. That makes Go a googol times more complex than chess. This was a surprising event as experts have only predicted this to happen only in another 10 years time.
One of the key to IoT adoption is making sense of data collected from a network of sensors & also from past historical databases. With AI technology, these data can be analysed extensively & “learned” to train machines to predict events or outcomes before it happens creating value or results without any human intervention.