Google Duplex might look cool, but don’t believe the hype

The Google Duplex demo was certainly eye-catching, but a year ago the world was getting similarly excited about the company's ill-fated Pixel Buds

Google Duplex represents the tricky next step for AI personal assistants: making them personal. In a headline-grabbing demo at its I/O conference, Google showed a system that sounded just as human as a human, littering its sentences with ‘ums’ and ‘ahs’ and cheesy upwards inflections. It was, perhaps, a glimpse of our near future. Or maybe not.

A year ago, at I/O 2017, the world’s press collapsed into a flurry of hyperbole at Google’s Pixel Buds. The promise of instant, on-the-fly language translation would “change the world”. A few months later, when Google released its world-changing product, the hype collapsed. “Overpromised, underdelivered”, lamented one review. But the promise was never (really) there.

The Pixel Buds could supposedly perform real-time translation between 40 languages. It was the Babel Fish made real. The reality was rather different. The Pixel Buds were a mess of a product born out of a (very good) product demonstration. The same will likely be true of Duplex. In our messy and muddled world – as Facebook found with its ill-fated M chatbot – this technology remains frustratingly limited.

A mistake oft-made when covering technology companies is to ask “what’s next?”, to always expect the Jobsian “one more thing”. The press (and consumers) demand it and, more often than not, the technology companies oblige. It is in this endless hype cycle that ideas like Duplex are imbued with undue prominence. Fool me twice, as the saying goes.

Perhaps the best example of this is Google Glass. At its I/O conference in 2012, Google sent jaws dropping with a live skydiving demo of its augmented reality glasses. But the actuality of (very much still work-in-progress) technology was somewhat different and, in January 2015, Google announced it would stop producing the Glass prototype. “We allowed, and sometimes even encouraged, too much attention to the program,” said Google X research lab head Astro Teller shortly afterwards. But hubris is addictive. (A Google Glass for factories was revealed in 2017).

As recent years have shown, technology firms would be better served fixing existing problems rather than creating more. From Russia’s attempts to use major technology platforms to subvert democracy to YouTube’s penchant for recommending videos filled with blood, suicide and cannibalism to children, Google has a long to-do list.

Read more: Google I/O 2018: here's all the big news you might have missed

Google and its contemporaries (well, mostly Facebook) will dangle the promise of AI as a solution to such ills. Don’t want your kids to be exposed to videos of Minnie Mouse being decapitated by an escalator? There’s an AI for that. Don’t want Google to make money from adverts promoting Holocaust denial? There’s a “scaleable algorithmic approach” to fix that.

Arguably of greater significance was Google’s decision to suspend all advertising around Ireland’s Eighth Amendment referendum. The May 25 vote on abortion rights has been targeted by a shady campaign of fake news and misinformation, with much of it boosted through online advertising on Google and Facebook. But such announcements don’t get tongues wagging. When you’re after positive headlines, pivot to sci-fi.

Enter Duplex and a raft of headlines praising Google for changing the world. Again. For a company that should really be focussed on fixing major problems with how it is impacting our lives and democracies, it’s a wonderful distraction. For the technology commentariat, it’s an opportunity to gaze glassy-eyed into the future.

Cynicism aside, Google is right to be putting limited resources into such projects. Duplex, which has been in the works for years, is likely a result of the company’s 20 per cent initiative (though Google hasn’t confirmed this) and remains a fairly small but crucial part of its overall ambition to make Assistant less clunky and more intuitive.

But for the time being, Duplex promises to fix problems – booking haircuts and making restaurant reservations – that are trivial at best. Its real promise – of being able hold a huge range of open-ended conversations in a human-like way – remains a decades–away dream reliant on an artificial general intelligence breakthrough. And if anyone is capable of making that breakthrough, it’s probably Google.

Duplex also has the feel of an idea created in an echo chamber. The two leads on the project, Yaniv Leviathan and Yossi Matias, are both men with long and distinguished careers at Google’s R&D centre in Israel. Because hey, wouldn’t it be cool if an artificial intelligence could do all that dull work you normally ask your real-life, human assistant to do? Yes, but it would also be cool if you could stop confusing black people with gorillas.

This article was originally published by WIRED UK