Updated 5 months, 2 weeks ago

The hype-and hope-of Artificial Intelligence

Last month, on his HBO show “Last Week Tonight,” John Oliver skewered media companies’ desperate search for clicks. Like many of his bits, it became a viral phenomenon, clocking in at nearly six million views on YouTube. At around the ten-minute mark, Oliver took his verbal bat to the knees of Tronc, the new name for Tribune Publishing Company, and its parody-worthy promotional video, in which a robotic spokeswoman describes the journalistic benefits of artificial intelligence, as a string section swells underneath.

Tronc is not the only company to enthusiastically embrace the term “artificial intelligence.” A.I. is hot, and every company worth its stock price is talking about how this magical potion will change everything. Even Macy’s recently announced that it was testing an I.B.M. artificial-intelligence tool in ten of its department stores, in order to bring back customers who are abandoning traditional retail in favor of online shopping.

Much like “the cloud,” “big data,” and “machine learning” before it, the term “artificial intelligence” has been hijacked by marketers and advertising copywriters. A lot of what people are calling “artificial intelligence” is really data analytics—in other words, business as usual. If the hype leaves you asking “What is A.I., really?,” don’t worry, you’re not alone. I asked various experts to define the term and got different answers. The only thing they all seem to agree on is that artificial intelligence is a set of technologies that try to imitate or augment human intelligence. To me, the emphasis is on augmentation, in which intelligent software helps us interact and deal with the increasingly digital world we live in.

Three decades ago, I read newspapers, wrote on an electric typewriter, and watched a handful of television channels. Today, I have streaming video from Netflix, Amazon, HBO, and other places, and I’m sometimes paralyzed by the choices. It is becoming harder for us to stay on top of the onslaught—e-mails, messages, appointments, alerts. Augmented intelligence offers the possibility of winnowing an increasing number of inputs and options in a way that humans can’t manage without a helping hand.

Computers in general, and software in particular, are much more difficult than other kinds of technology for most people to grok, and they overwhelm us with a sense of mystery. There was a time when you would record a letter or a document on a dictaphone and someone would transcribe it for you. A human was making the voice-to-text conversion with the help of a machine. Today, you can speak into your iPhone and it will transcribe your messages itself. If people could have seen our current voice-to-text capabilities fifty years ago, it would have looked as if technology had become sentient. Now it’s just a routine way to augment how we interact with the world. Kevin Kelly, the writer and futurist, whose most recent book is “The Inevitable: Understanding the 12 Technological Forces That Will Shape Our Future,” said, “What we can do now would be A.I. fifty years ago. What we can do in fifty years will not be called A.I.”

You don’t have to look up from Facebook to get his point. Before we had the Internet, we would either call or write to our friends, one at a time, and keep up with their lives. It was a slow process, and took a lot of effort and time to learn about each other. As a result, we had fewer interactions—there was a cost attached to making long-distance phone calls and a time commitment attached to writing letters. With the advent of the Internet, e-mail emerged as a way to facilitate and speed up those interactions. Facebook did one better—it turned your address book into a hub, allowing you to simultaneously stay in touch with hundreds, even thousands, of friends. The algorithm allows us to maintain more relationships with much less effort at almost no cost.

Michelle Zhou spent over a decade and a half at I.B.M. Research and I.B.M. Watson Group before leaving to become a co-founder of Juji, a sentiment-analysis startup. An expert in a field where artificial intelligence and human-computer interaction intersect, Zhou breaks down A.I. into three stages. The first is recognition intelligence, in which algorithms running on ever more powerful computers can recognize patterns and glean topics from blocks of text, or perhaps even derive the meaning of a whole document from a few sentences. The second stage is cognitive intelligence, in which machines can go beyond pattern recognition and start making inferences from data. The third stage will be reached only when we can create virtual human beings, who can think, act, and behave as humans do.

We are a long way from creating virtual human beings. Despite what you read in the media, no technology is perfect, and the most valuable function of A.I. lies in augmenting human intelligence. To even reach that point, we need to train computers to mimic humans. An April, 2016, story in Bloomberg Business provided a good example. It described how companies that provide automated A.I. personal assistants (of the sort that arrange schedules or help with online shopping) had hired human “trainers” to check and evaluate the A.I. assistants’ responses before they were sent out. “It’s ironic that we define artificial intelligence with respect to its ability to replicate human intelligence,” said Sean Gourley, the founder of Primer, a data-analytics company, and an expert on deriving intelligence from large data sets with the help of algorithms.

Whether it is Spotify or Netflix or a new generation of A.I. chat bots, all of these tools rely on humans themselves to provide the data. When we listen to songs, put them on playlists, and share them with others, we are sending vital signals to Spotify that train its algorithms not only to discover what we might like but also to predict hits.

Even the much talked-about “computer vision” has become effective only because humans have uploaded billions of photos and tagged them with metadata to give those photos context. Increasingly powerful computers can scan through these photos and find patterns and meaning. Similarly, Google can use billions of voice samples it has collected over the years to build a smart system that understands accents and nuances, which make its voice-based search function possible.

Using Zhou’s three stages as a yardstick, we are only in the “recognition intelligence” phase—today’s computers use deep learning to discover patterns faster and better. It’s true, however, that some companies are working on technologies that can be used for inferring meanings, which would be the next step. “It does not matter whether we will end up at stage 3,” Zhou wrote to me in an e-mail. “I’m still a big fan of man-machine symbiosis, where computers do the best they can (that is being consistent, objective, precise), and humans do our best (creative, imprecise but adaptive).” For a few more decades, at least, humans will continue to train computers to mimic us. And, in the meantime, we’re going to have to deal with the hyperbole surrounding A.I.

The New Yorker | By Om Malik