AI, the new electricity

Two innocuous-looking letters got together to dominate technology in 2018. It was the year we discovered just how disruptive, powerful and revolutionary AI (artificial intelligence) will be, giving the human race — its inventor — a run for its money.

Some of the most startling AI came from Google’s DeepMind, which showed how neural networks could beat humans at what they thought they alone could do best.

All through the year, practically every week, there were announcements about some new hitherto unimagined achievements with AI and machine learning at their core. Machines learned to teach themselves with no human supervision, they started thinking like humans, even taking over creativity in fields considered essentially human — music, painting, writing and debating. Just this week it was found that AI can teach itself to cheat, even hiding the skill for subsequent use.

But AI didn’t come suddenly out of nowhere, and now that we’re in 2019, it isn’t going into the background to make way for some other hyped technology to take its spotlight. Through this year, we’ll certainly hear a lot about 5G, which is needed to carry so many AI-based solutions, but this won’t carpet the planet as quickly as desired, and will not impact India for a while yet.

Elsewhere in the world there’s more of an urgency to roll out 5G connectivity because it will fuel driverless cars. In India, the hope is that 5G will help deliver healthcare to remote areas, but the infrastructure isn’t in place yet.

After the number of stunning AI achievements in the past few months, the frenzied pace of innovation is only likely to accelerate through 2019. The year gone by was about breakthroughs that happened at too breakneck a speed for real-life uses to become apparent. The new year will be about AI becoming like electricity, powering critical everyday things. Just as we stopped gawking at electricity, we will gradually move to not gushing about AI because it will be difficult to separate it from other aspects of living. Industries, institutions and, more frighteningly, governments and politicians will explore its uses, as they have already begun to.

The everyday consumer has been introduced to AI through the smartphone, with companies claiming to have used the technology to make everything from photography to battery life, where virtual assistants are supposed to have made routine tasks easier, and where the gadget learns the user’s behaviour and makes all manner of adjustments specific to the person. With marketing being the main objective, companies have all but deified personalisation to the point where privacy has been torn down and personal choices reduced because people gradually take less and less trouble to look for anything beyond what is put in front of them. This narrowing down of choices will become evident as there are more calls for AI to be transparent and ethical. But just as has happened with Facebook losing face over its use of data, so it will be difficult with other companies to reverse the damage already done to privacy and the exercise of free unbiased personal choice.

Let’s say I’m thinking of buying a short skirt. Because I’ve looked for, and perhaps bought long skirts in the recent past, I will find ‘recommendations’ for long skirts everywhere I turn. Encountering search results and ads on every app and online resource, I will have to work harder to get past the data being thrown at me and get to non-personalised options.

From here on, the potential impact of AI, specially when combined with robotics and other related fields, will also begin to be realised and be a source of worry. The biggest of these is how the fake mixes in with the real in everyday life. This is getting to be so pervasive that the very concept of trusting what you see and hear to be authentic is on the verge of being destroyed. Quite apart from fake news, generative AI has shown how it can fake voices, create incredibly real images of faces of people who don’t even exist, and even fake fingerprints by creating a skeleton key master print from going through thousands of prints at high speed to come up with a match. On the website of a start-up called Lyrebird, you can register and proceed to record voice samples and, after a few hours, come up with an authentic-sounding replica of your voice. Whatever is typed in and fed to the voice-avatar will now sound startlingly close to your own voice — the more you train the avatar, the more authentic it will sound. Fakery will be one of the most unwieldy challenges to deal with from now on.

COMMENT NOW