The 2018 edition of Google I/O, the search giant’s annual developer conference, shifted the limelight to the big new area of focus from the company: Artificial Intelligence (AI) and Machine Learning (ML). If there was one theme underlying most of Google’s keynote, it was that AI and ML advancements from the company were going to drive most of its products from Android or the self-driven Waymo car. The biggest highlight of the event was no doubt the demonstration of Google Assistant using the company’s Duplex technology to make an actual call to an unsuspecting human. What was remarkable and also eery, was that Google Assistant was able to pass itself off as human to the caller as it carried out the task.
While Google Assistant will soon make calls on behalf of users in the US, its capabilities will be limited to just three tasks: booking a restaurant, a hair salon appointment and checking closing/holiday hours for a business. Nonetheless, the Assistant’s new prowess will raise questions on the future of this technology and whether we might soon have virtual assistants that sound more human.
While Google Assistant’s voice-calling ability is still an experiment, according to the company, it was evident from the keynote that digital assistants which can carry out our day-to-day tasks while sounding as ‘natural’ as possible are no longer just the stuff of science fiction.
The other highlight of Google I/O was Android P, which again has a layer of AI and ML built-into it. This is what is driving ‘Actions’ where Android P will predict the kind of action a user might want to carry out next. For instance, plugging in headphones could prompt your Android phone to ask if it should open a music app like Spotify, depending on one’s usage habits.
Then there’s Adaptive Battery, which figures out your app usage and will prioritise battery for the ones which are used most, to help improve efficiency and extend the smartphone’s average battery life. For this, Android is partnering with DeepMind, which is an Artificial Intelligence company owned by Google’s parent Alphabet. For the Adaptive battery, Android relies on a deep neural network to predict which apps are used more.
Google will also have a new Machine Learning ML Kit for app developers, which will work on both Android and iOS. Again, the aim to help developers make apps which use these ML Kit tools to aid with “text recognition, face detection, image labeling,” etc.
The other product that’s getting a touch of AI and ML is Google News, which has been completely redesigned from scratch. Google News will now be the only app for news from the company, and integrates the company’s earlier Newsstand app as well. Google News will rely on a “new set of AI techniques,” according to the company. These will look at the flow of news information, analyse it and then sort it into storylines.
According to the company, it also means that Google News will now understand “the people, places and things involved in a story as it evolves, and connects how they relate to one another”. While Google insists it does not intend to make editorial calls or judgements around news organisation, its new product will rely on AI to sort the stories and then link them together to present a picture to the user.
Then there was the announcement around Gmail which will now have a ‘Smart Compose’ feature, again powered by AI, ready to complete your sentence before you even type them in, and understanding the context of your email. For instance, based on the subject of the email and the first half of the sentences, ‘Smart Compose’ will predict text to fill in.
Google’s research has also shown that their AI model can predict cardiovascular events more accurately. Using these tools doctors could potentially find health problems early on and initiate a better diagnosis. In a Google study published in the Nature Research journal, the system made predictions regarding patients who had been hospitalised and whether there was likely chance of them being admitted again due to another episode. As Google’s study notes, this does not meant that AI will replace doctors in the future. Rather it could aid doctors and make healthcare more efficient in the future.
Disclaimer: The reporter is in Mountain View, California, attending Google I/O 2018 at the invite of Google India.