AI will go where the data is

AI will go where the data is

When discussing artificial intelligence, people often end up visualizing what our world will look like 50 years from now. Let’s not look that far, let’s instead focus on which industries AI is most likely to revolutionize in the coming few years.

The key here is data. AI is nothing without a good amount of intelligence (it’s in the name, duh). For AI to be effective, the data collection needs to work well, and it must be available.
Because of this, AI might actually be most effective in some of the world’s oldest industries, like agriculture and health care.

In agriculture, the vast amount of harvest data is used to place seeds and chemicals exactly where they’re needed, autonomous harvesting, automatically watering the soil, and notifying the farmer of what’s going on with the crops. With sensors in the soil monitoring temperature, water levels, and movement, farmers can water only the areas that actually need it, which is both more efficient, cheap, and environmentally friendly.

In healthcare, AI helps doctors make better decisions for each patient with predictive analysis, as well as identify at-risk patients by using pattern recognition. AI has also improved the chances of early detection of certain diseases like breast cancer. With AI, mammograms can be reviewed 30 times faster with a 99 percent accuracy, which can drastically reduce the number of unnecessary biopsies.

Within this field, it also helps that many consumers are joining in on the trend, using wearable devices powered by the Internet of Medical Things (IoMT). The data these gadgets register can be used to help consumers change unhealthy habits, facilitate for doctors to detect and monitor heart problems earlier, and give medical staff better insights into day-to-day patterns.
A more controversial usage of AI in healthcare is humanoid robot caretakers for the elderly. These robots could decrease the pressure on the healthcare system by treating each patient in an optimal way and offering human-like interactions (or even conversations) which could help patients with diseases like dementia stay sharper. Agriculture and healthcare are not the only sectors at the forefront of the AI revolution. Other industries that already have enough data for AI include transportation, delivery services, manufacturing, and construction.

Stop typing, start talking

Stop typing, start talking

Machine learning and AI are fundamentally changing the way we interact with our computers. Perhaps the best interface will be no interface. Let’s just talk.

“It’s part of the human condition to think that if we struggle to use something, we assume that the problem resides with us,” said Jonathan Ive, Apple’s chief design officer. The best type of user interfaces are the simplest ones, the ones that work intuitively and doesn’t require much analysis on our parts, but adapt to our ever-changing needs. This insight uncovers a hidden reality of using computers: we have to adapt to their behavior. We learn their foibles, they don’t learn ours. But perhaps we’re getting closer to the ideal user experience – no interface at all. Chatbots and voice are still at their very beginning. But everything points towards that we will be talking a lot more in the future.

“We are going from talking through messaging apps to chatting to machines”

Computing paradigms change every 10 to 15 years; they’re typically defined by how they operate with the outside world – meaning we have to change with them. The first computers purely operated via command-line (or text) input. They required linguistic skills of a precision that the Academie Française would have been proud of. The graphical interface (GUI), pioneered by the Xerox Alto, popularized by the Mac and dominated by Microsoft Windows, took hold in the late 1980s. GUIs were more forgiving visualizing everyday metaphors like files and folders on a color screen. This is the computing most of us know. Multi-touch computing, pioneered by the Apple Iphone, was a third revolution, point and draw with your finger, what could be simpler? The Iphone fundamentally changed the way we interacted with technology, our expectations, because the whole screen became a playing field.

Chat is natural

Smartphones, in turn, paved the way for the rise of messaging apps. We now have countless ways of contacting each other, whether it’s on Imessage, Whatsapp, Messenger, Slack, Skype or Wechat. And since it makes sense for companies to try to talk to us, using the same channels we use to talk to one another – chat and chatbots have received a lot of hype – becoming, you might say, our latest interface. The reasons are clear: chat is natural and we spend a lot of time in chat applications. Turns out chatbots are also ludicrously easy to build. But it also turns out, building a great chatbot is a lot tougher than building a chatbot. If you’ve ever tried chatting with a chatbot you’ll know why; the conversation is dull and repetitive. God forbid you ask an original question only to be met with utter incomprehension.

So, we’re still pretty far from the ultimate interface, but no doubt, things are happening. Today, the technology is converging and leaps made in one field serve another. Natural language processing (NLP) enables chatbots, image recognition enables self-driving cars, voice recognition enables Alexa, Google Home, Siri. Those are all different branches of machine learning and we’re getting better and smarter at it, at an increasingly faster rate. A few companies are now starting to reach that level but we’re still in the early days. Yet, according to experts, by 2020, 85 percent of all customer interactions won’t require human customer service reps; indeed, those interactions will happen over chat, but also over voice.

We are going from talking to one another through messaging apps to chatting to ­machines. What’s the next step? Eliminate typing, and use your voice. Going back to the point on the importance of keeping user interfaces simple, voice is a big deal. To quote the epo­nymous book, the best UI is no UI. No design is required if you could simply talk to your device.

Today, voice AI such as Siri or Alexa, are limited by two things: technology and architecture. On the technology front, speech recognition and text recognition still have a lot of room for improvement, especially if your English is somewhat accented. (Fun experiment, ask Siri to “Google Tchaikovsky” for you with a French accent, you’ll get surprising results.) Their architecture is based on general themes, the AI is able to draw context from the user’s request, classify it and answer it accordingly. What it has a hard time doing however, is to follow a conversation, remember pieces of information mentioned three questions back and use it when needed. There’s no dropping birthday gifts hints with Alexa.  But thanks to the millions of users that interact with it regularly, the AI is getting plenty of training and gradually getting better.

A voice AI good enough for us to freely chat with would be extremely liberating: no more staring at your screen constantly, just chat with your AI, how cool does that sound? Nevertheless, voice AI raises some really challenging UX problems. How do you teach your users to use an interface which is actually invisible?  What will be the standard keywords to which Voice AI will respond to and who will set them?

We know the world is changing

Can Voice ever be good enough to be totally unscripted, feel as seamless as talking to a fellow human? The answer to this question is more a matter of belief, than it is hard science. We cannot anticipate the changes that will happen with the exponential development in tech and what we will be able to do. For now, a “Her”-like society is definitely science fiction.

What is very real, however, is the short-term impact voice and chatbots will have on the way businesses interact with their customers. Indeed, 32 percent of executives say voice is the most widely used AI technology in their business. Six billion connected devices will proactively ask for support by 2018. By the end of 2018, customer digital assistants will recognize customers by face and voice across channels and partners. HSBC has already implemented voice recognition as a secure access to one’s banking details.

We all know that the world is changing and it’s changing faster than ever. Not that long ago we were all going nuts about tactile screens – “it works without buttons!” – and now we live in a time in which soon all homes in developed countries will be equipped with voice AI devices to facilitate and organize our lives. And where businesses will interact with their customers in a way that is barely invented yet.

AI will power our life

AI will power our life

There’s a bigger picture beyond the progress of Artificial Intelligence. Just like when electricity arrived we can’t yet see the full effect. And just like with electricity – life without AI will be inconvenient and unpleasant.

In October 1881 in the small British town of Godalming, the world’s first public electricity supply went on stream, lighting a few dozens incandescent lamps in the town’s streets. The small  local firm, Calder and Barrett, had, with their small hydroelectric generator, passed a milestone in the electricity revolution. From these humble beginnings, electricity rapidly became part of the fabric of our lives. The American inventor, Thomas Alva Edison, opened his first public generator in London some three months later, followed by a New York generator some nine months after that.

“Entire industries will benefit from the arrival of the AI, as they did with electricity”

In the US the process of electrifying the nation was mostly complete within 50 years. Today electricity is a utility, not even a commodity. It just works. You can’t really imagine life without it. It is the same everywhere in the world. Everything we do, we buy, is based on the assumption that we’ll have access to electricity. And today, more than 5.5 billion of us do. So when Andrew Ng, one of the most influential forces in today’s artificial intelligence boom, describes AI as the new electricity we stop and take notice. After all, electricity changed industries, jobs, our everyday lives and our social and domestic relationships. Could we perhaps glimpse the second-order effects of artificial intelligence by understanding how electricity shocked the world?

The first experiments with electricity showed the technology could work, lighting bulbs in daylight is a scientist’s a-ha moment. But it wasn’t a study of electricity itself that wrought changes, it was the application of electricity that did. What is interesting is using electricity to power lights to extend the practical day. Or using electricity to reduce the hardship of domestic chores. Or using it to power new classes of production processes – like The Haber-Bosch process for fixing atmospheric nitrogen and creating artificial fertiliser is heavily dependent on electrical energy. Many of today’s AI a-ha moments are a scientist’s Eureka: being able to transcribe speech to written text or describe what is in an image at higher than human quality. But, like with electricity it is the applications, not the technology, that will be the impact of AI in our lives.

Life will seem inconvinient

AI will be deployed rapidly, made available to every part of the industry and home very quickly. In the UK, by 1933 one in three houses had electricity and a further ten years down the road two out of three houses were electrified. The spread of electricity was accompanied by a rapid increase in the intensity of use. The amount of electricity used per consumer grew from about 2 MwHs per household per annum in the 1930s to close to 11 MwHs per household by 2014. This was accompanied by a dramatic drop in the price, almost four-fold in real terms over that same period, and a drop of about 20 times from the turn of the 20th century to the turn of the 21st. AI will be deployed far more rapidly than the appliances that used electricity – because the components it needs to work are already in place. Smartphones in our pockets, wireless internet, digital cameras, cloud-based services. Artificial intelligence will be ubiquitous, wherever there is the internet (and in many places where there isn’t.)

AI services will become indispensable. Life will seem inconvenient, even unpleasant, without it. When I was a child, I would visit my grandparents’ home in Lahore, a house they had lived in since the partition of India in 1947. This old house didn’t have a flush toilet. A minor thing, humanity had lived without flushing loos most of its time on earth, but an inconvenience to make a Westernised grandson ill-humoured. Our expectations for interactions to be AI-powered will rise rapidly. We’ll expect our clinicians to spend time explaining our diagnoses to us, not diagnosing us, because AI systems will have more effectively identified our ailments than any human. We’ll tire of waiting in traffic queues with aching backs because our autonomous vehicles will find the best route for our journey.

New industries will rise

Entire industries will benefit from the arrival of the AI as they did with electricity. In truth the electrical companies did well, but not as well as the fossil fuel companies who provided much of the raw input. Nor as well as the cambrian ­explosion of firms who could exploit new business models because of the arrival of electricity. Shopping malls make no sense without electricity. Television makes no sense without electricity. Even frequent air travel is not possible without electricity.

AI will drive an analogous Schumpeterian process: industries arising, replacing others. Take the car industry: it has a century-old model of selling us cars through distributors who make money through service and repair. The car industry itself supports the advertising and media industry who are enlisted in manufacturing desire around this mode of transport.

Yet the car industry looks like it will be upended by new modes of on-demand transport rental: autonomous vehicles routed to us by algorithms. Much like electricity, the biggest impacts of AI will be felt outside its home industry. AI will also transform work and with that a transformation of our social relationships. Rural electrification in America, which occurred in the years before the depression, was driven by women. Unlike men, who worked seasonally, women worked from dawn till dusk throughout the year. Electricity, with its washing machine and kitchen stove, alleviated the burdens of their work and, with its radio, connected them to wider society.

How will AI transform our labor structures and with that our wider society? Electricity provided power where it was needed, automating manual tasks, for example washing clothes. It also provided lighting, extending the working day and creating time for in-home leisure. Radio and then television appeared to fill the gap. Artificial intelligence will free up time by taking up some of our cognitive load. It could create free time for us, the same way cheap lighting enabled by electrification extending the day, for both work and leisure. There are any number of taxing but low-value tasks that you could foresee leaving to machines. What we do with all that time, remains to be seen. No wonder that as I talk to business leaders in industries like retail, health diagnostics, professional services and finance, their number one priority is artificial intelligence.

An engaged public debate

Of course, electricity brought with it fears of this ‘mysterious fluid’. American President, Benjamin Harrison had the White House wired for electricity but refused to touch the switches for fears of shocks. It probably didn’t help that it was during Harrison’s term, the first man was executed in the electric chair. William Kemmler died in August 1890, before Edison’s Pearl St generator opened. His execution was botched and resulted in a gruesome 8-minute ordeal. Like electricity, artificial intelligence is bringing forward many fears. Will AI amplify the biases in the world? Will AI lead to persistent and chronic unemployment? Will AI create new megalithic dominant firms controlling large parts of the economy?

These are all real concerns. AI systems have already been demonstrated to systematically bias the bail assessments of black prisoners in the US. And academics, Daron Acemoglu and Pascal Restrepo, have demonstrated that the use of robots in industry depresses both wages and employment levels. And one only need to look at the dominance of Amazon in retail, and Google and Facebook in advertising, to see the risks of market dominance driven by data monopolies. The only way to manage these fears is to have an engaged public debate on the many ways AI will impact the economy, something that is happening in many countries today. For example, in the UK both houses of Parliament have public consultations on the impact of AI underway.

When we think about AI today, we need to go back perhaps not to Edison’s pumping station on 1892. Perhaps AI today is more like electricity in the 1920s America, spreading fast but in limited intensity. But around that spark of intelligence, entrepreneurs and incumbents are figuring out how to apply this soon to be ubiquitous technology in our everyday lives. I believe AI services will generally have one of four major benefits: Relate, not diagnose – AI does diagnosis saving time for human communication; Maintain, not repair – advanced diagnostic systems spot likely failures well ahead of time allowing proactive maintenance; Connect, not collect – machine learning systems end up personalizing specific needs and requirements rather than broad buckets; Take out the boredom – machine learning systems do the boring work, leaving humans to do the interesting, challenging parts.

Four benefits from AI

Relate, not diagnose: Online test-prep company Magoosh uses AI software for customer service, allowing the human providers to respond to more test-takers more quickly: the software has reduced Magoosh’s queue of customer requests by half, and it has made her team’s goal of responding to every customer within 24 hours more manageable. Dr Rajesh Jena, a consultant neuro-­oncologist at the University of Cambridge Cancer Centre, has worked with Microsoft researchers to develop a tool which provides accurate 3D visualizations of tumors and organs. This has cut down the time it takes to identify a neuroblastoma from a few hours to four minutes, allowing the specialist more time to counsel the patient.

Maintain, not repair: As our devices get smarter and connect to the net, the so-called internet of things, we’ll be able to proactively maintain equipment rather than repair it when it breaks. One favorite area is to detect faults in offshore wind-turbines weeks before they become critical. This saves money, as prevention is better than cure. It isn’t just wind turbines were prevention is better than cure. The same goes for humans. Cardiogram, an AI-enabled app on Apple Watch, detects irregular heartbeat with 97 percent accuracy, while NVIDIA’s efforts to combine 3D modeling with AI could spare 60 percent of patients from getting angiogram, an invasive and costly scan.

Connect, not collect: One of the leading insurance and investment companies in the U.S., Transamerica, employs machine learning to make product recommendations to its potential customers. McGraw-Hill Education presented its web-based artificial intelligent assessment and learning tool last year, using graph theory to test the process and speed of learning for each student. Machine learning algorithms can follow students’ learning patterns and create personalized learning pathways for individual learners.

Take out the boredom: Automated systems already take the tedium out of flying a plane. Pilots are left to handle the take offs and landings. Autonomous trucks, buses and cars could take out the tedium of driving for long stretches, leaving humans to manage passengers – or even relax. AI systems in legal document discovery spare junior lawyers hundreds of hours of combing for juicy morsels. Those lawyers can be put to more valuable tasks.