Artificial intelligence isn’t the future. It’s already here.

AI is in our pockets, through smartphones that can understand voice commands and the flick of a finger. You can find it in computers that can match your resume to a job posting or your dating profile to your soul mate. And it’s in chat-bots that navigate social media feeds and driving apps that guide you on your commute home.

“We are at a place today where it’s much easier to answer the question: Where isn’t it going?” says Premkumar Natarajan, Michael Keston executive director of the USC Information Sciences Institute.

It has even attained the role of personal stylist, Natarajan says. “Who would have thought, a few years ago, that you would stand in front of a TV with a tiny camera that captures you in high-res, and ask it, ‘Which clothes should I wear?’”

AI has been around for a long time, but it was found mostly in simple technology like a car’s anti-lock braking system or a coffee maker set to a timer. Now, artificial intelligence takes on more complex tasks like rating fashion or brewing coffee when it senses daylight, Natarajan says.

While many people attribute AI’s growth to technical advances, Natarajan credits the momentum to information availability. “When you have lots and lots of data, you can design algorithms to learn from massive amounts of data,” he says. “The thing that has enabled us to design things is the explosion in the field of deep learning: a class of algorithms that exhibit essentially unlimited capacity to learn from more and more data.”

Those two complementary phenomena—data and algorithms that learn—are driving a future in which AI is part of everyday life. “If the world is a sponge, AI is water,” Natarajan says. “It’s dripping slowly and occupying every vacant space of that sponge.”

While AI is here today, in the present, it may also forecast what’s to come. Natarajan studies predictive analytics—in other words, he uses technology to look into the future. “I know, it sounds like computational astrology,” he laughs. But deep data tools that can analyze social media feeds could potentially predict whether governments will fall. Measuring hospital parking lot flow could assess whether a flu epidemic will peak in your zip code.

At USC, researchers like Natarajan are working on dozens of projects in areas where artificial intelligence has been under study for decades—like language—and where the tools are just starting to make forays—such as efforts to combat human trafficking or diagnose autism.

TALKING TECH

Anyone who has used the Google Translate app, which can instantly decode words from more than 100 languages, is a beneficiary of the work of USC’s Kevin Knight.

Knight, Dean’s Professor of Computer Science at the USC Viterbi School of Engineering, has worked in automatic language translation and machine translation since the early 1990s.

He’s fascinated by tools that can seamlessly bridge the divide between languages. While most of the work has focused on translation of languages to English, Knight is interested in expanding across more languages. “There are 4,000 languages on Earth, and for many of them, translation just doesn’t exist,” he says.

This language barrier especially matters in times of disaster. Agencies and nonprofits must prioritize humanitarian assistance based on need. “If we can’t understand the tweets and the blogs, it makes it hard to organize,” Knight says.

But there’s another reason to tackle machine translation: opening up the world to everyone. “You should be able to go anywhere in the world and, wearing augmented reality glasses, you should be able to read signs in your own language,” Knight says. “I should be able to speak to anyone on the phone, anywhere in the world.”

Artificial intelligence is also reaching over into a creative side of communication called natural language generation: computers’ ability to talk or write in a conversational, human way. Knight’s research group designed algorithms that can generate original poetry, and they hope one day to create stories and even movie scripts as algorithms grow more sophisticated in analyzing style and prose. In a 2016 contest for computer-generated poetry, Knight’s group’s entry won first prize.

In the future, Knight says you may be able to order a movie or story that’s completely customized to you. But beyond creating independent works, artificially intelligent machines could assist humans in their own creative pursuits, like helping an author overcome writer’s block by offering suggestions of new directions to pursue.

The technology is progressing quickly, but engineers still haven’t entirely figured out how to get machines to write creatively to a human standard. “Four years ago, we were still struggling to get the subject to agree with the verb,” Knight says. “Now machines can read human works and produce similar sorts of documents. The challenge is to get it to where you’ve got a beginning, a middle and an end to a story. These things have a topic, but they tend to wander.”

Translation and poetry may also usher in a new era of how humans interact with computers. Work on so-called virtual humans would mean that no one has to learn an operating system ever again, says William Swartout, a USC Viterbi research professor of computer science and chief technology officer at the USC Institute for Creative Technologies.

ILLUSTRATIONS BY CHRIS GASH

The institute is creating virtual humans that are more dynamic than just a voice like iPhone’s Siri or Echo’s Alexa, Swartout says. They’re visible, embodied characters, like avatars. “Because they have bodies, they can not only use their voices like natural speech, but also use body language, which opens up all kinds of capabilities,” he says. In particular, virtual humans can pick up on body language like eye contact and gestures, which researchers know are important to build rapport. Studies have shown that people can feel more comfortable talking about sensitive topics with a virtual human than a real person, Swartout adds.

Virtual humans have applications in training, teaching and therapy, but they also represent a new way of communicating with machines. “Everyone is familiar with the desktop metaphor that has been around for 30 years now, and it has been successful,” Swartout says of computer operating systems. “But I actually think virtual humans are a new metaphor for interaction. Instead of interacting with a screen, you’re interacting with a character who can give advice and be empathetic.”

That means that in the future, dealing with a computer could be as simple as talking to it as if it were a person. The computer, as a virtual human, could bring social factors to the interaction, for example: A computerized tutor, for example, could convey not only knowledge, but also a passion for its topic.

FIGHTING CRIME

Artificial intelligence is also moving into law enforcement. One project at USC works to combat human trafficking by pulling vast quantities of information from internet forums, chats and online databases.

Software called Domain-specific Insight Graphs trawls the open web and dark web to give law enforcement officers valuable clues left behind by criminals. It helps them search for children and teens who might be trapped in the escort industry. Unearthing information like a phone number, location, alias and photos can be the break that law enforcement needs to locate human trafficking victims.

Craig Knoblock, a USC Viterbi research professor of computer science, says the technology is powerful because it can sift through around 200 million pages of data, and the amount of data is constantly increasing. “Before this, law enforcement agencies would typically have to go and do Google searches,” he says. “Those searches lacked a history, so you can’t see what was there last week or last year.”

The software is part of a system developed at USC and funded by the Defense Advanced Research Projects Agency. More than 400 law enforcement groups have signed up to get access to the software, and the researchers have received reports that it has significantly improved agents’ ability to find and prosecute traffickers.

HEALTHIER HORIZONS

When it comes to diagnosing medical problems or understanding conditions, artificial intelligence offers a new way to see beneath the surface. Work by engineer-scientist Shri Narayanan, USC professor of electrical engineering, computer science, linguistics, psychology and pediatrics, is especially illuminating.

Narayanan, who holds the Niki and C. L. Max Nikias Chair in Engineering, uses machine learning and data science to help health professionals diagnose developmental disorders, as well as to guide the treatment of mental health issues. One project collects data on children’s smiles and laughter with the goal of creating tools to identify early signs of autism. Laughter is just one of the many non-verbal ways that humans use vocally to socialize (a category that includes sighs, pauses and affirmations like “mm-hmm”). People with autism often miss the timing and context of these nuanced non-verbal cues. By analyzing when and why people laugh and smile, Narayanan aims to develop markers for identifying impaired communication.

With more sophisticated mathematical models for emotions, scientists could also develop computers capable of natural, human-like conversations and interactions. Talking robots that fool you into thinking they’re human are still a long way off, but Narayanan already sees huge leaps in the ability to measure and find patterns in data.

ILLUSTRATIONS BY CHRIS GASH

AI can also detect unconscious biases. In a recent project, Narayanan’s group used special language-dissecting tools to study characters in nearly 1,000 film scripts. They confirmed many assumptions about gender differences in the film industry: Female characters are younger, tend to speak less, and use less language about achievement than male characters.

“Using artificial intelligence, we can see the patterns that are unfolding unconsciously, and we can pick up on things you wouldn’t necessarily notice just watching the movie,” Narayanan says.

Artificial intelligence is also opening up avenues to address persistent social problems. That’s the domain of Milind Tambe, co-director of the USC Center for Artificial Intelligence in Society, where predictive analytics is tackling some big challenges.

One project in Africa calculates where big-game poachers will strike, and also suggests places where rangers should conduct strategic, randomized patrols to thwart illegal hunting. It may seem unrelated on the surface, but the same sort of thinking can help people better understand how HIV spreads in a community.

“These are challenges where currently humans are in charge, whether it’s how to spread info about HIV among homeless youth, or how to stop poachers, or how to most effectively use our limited resources to prevent terrorist attacks,” Tambe says. Using artificial intelligence to address these challenges, he adds, “is a win-win-win: great for solving the problem, for research and for students.”

Can We Trust AI?

From Westworld to Terminator, the fictional world is full of murky ethics surrounding artificial intelligence. But USC researchers who create artificial intelligence systems don’t worry that these systems will start to attack humans—at least not anytime soon.

“At the moment, I don’t think AI poses more of a threat than any other computing technology, but in the long run it can,” says Paul Rosenbloom, director for cognitive architecture research at the USC Institute for Creative Technologies. “At some point, we have to view AI systems as if they are beings—maybe like animals, or perhaps like people—and from there, we have to figure out the rights and responsibilities that need to be assigned to them.”

Our society makes different rules for children and adults—for example, young people can’t vote before they are 18 and can get a driver’s license before 16, Rosenbloom points out. In the same way, future societies may apply differing rules for AI systems.

Rosenbloom also wonders what would motivate an AI system to do good in the world. “Some motivations that people have, like honor and responsibility, we need to carry over. But others, like having a family, don’t necessarily carry over,” he says. By understanding and imbuing AI with the right set of motivations, scientists could perhaps bake morality into AI.

Shri Narayanan, USC professor of electrical engineering, computer science, linguistics, neuroscience, psychology and pediatrics, says that since machines were created, people have always wrestled with a central question: How much do we trust them?

Despite the concerns, Narayanan focuses on the good that AI could do for people around the globe. “I hope that collectively as a society we are mindful about how we build and use things,” he says. “This is true of all instruments: fire, mechanical tools, everything. This is a debate that we have had for a long time and will continue to have.”

Science/Technology Artificial Intelligence Emerging Technology Information Sciences Institute Innovation Research

Trojan Family Magazine

The Future of Artificial Intelligence is Here — in Your Pocket and at USC