Artificial intelligence is shaping up as the next industrial revolution, poised to rapidly reinvent business, the global economy and how people work and interact with each other.
Andrew Ng, chief scientist at Chinese internet giant Baidu Inc. and co-founder of education startup Coursera, and Neil Jacobstein, chair of the artificial intelligence and robotics department at Silicon Valley think tank Singularity University, sat down with The Wall Street Journal’s Scott Austin to discuss AI’s opportunities and challenges. Here are edited excerpts.
MR. AUSTIN: Andrew, there’s a lot going on with artificial intelligence. What is Baidu focused on?
MR. NG: For large enterprises like Baidu, AI creates two big pockets of opportunities. One is our core business. Web search, advertising—all of that is powered by AI today. For example, we run a very large food-delivery service, and when you order food, we use AI to predict how long the food will take to get to you. That includes deciding which motorcyclist to dispatch to pick up the food, so they arrive just as the food is cooked and fresh, and they can get it to your door while it’s as hot as possible.
In addition to strengthening our core business, AI is creating a lot of new opportunities. Just as about 100 years ago electrification changed every single major industry, I think we’re in the phase where AI will change pretty much every major industry.
So part of my work at Baidu is to systematically explore new verticals. We have built up an autonomous driving unit. We have a conversational computer, similar to Amazon ’s Alexa and Google Home. And we’re systematically pursuing new industries where we think we can build an AI team to create and capture value.
MR. AUSTIN: Let’s talk about speech recognition. I believe someone in your program has said that the hope is to get to the point where it is 99% accurate. Where are you on that?
MR. NG: A couple of years ago, we started betting heavily on speech recognition because we felt that it was on the cusp of being so accurate that you would use it all the time. And the difference between speech recognition that is 95% accurate, which is where we were several years ago, versus 99% accuracy isn’t just an incremental improvement.
It’s the difference between you barely using it, like a couple of years ago, versus you using it all the time and not even thinking about it. At Baidu we have passed the knee of that adoption curve. Over the past year, we’ve seen about 100% year-to-year growth in the daily active use of speech recognition across our assets, and we project that this will continue to grow.
In a few years everyone will be using speech recognition. It will feel natural. You’ll soon forget what it was like before you could talk to computers.
MR. AUSTIN: Neil, at Singularity what sort of trends are you seeing from an AI perspective?
MR. JACOBSTEIN: Just since the beginning of 2017 we’ve seen a team at Northwestern develop an AI that could solve the Raven Progressive Matrices Test, an intelligence test of visual and analogical reasoning, better than the average American.
We also have seen a team at Imperial College London develop an AI that could diagnose pulmonary hypertension better than cardiologists typically do. Cardiologists have about 60% accuracy. This system does 80% accuracy. And in January of this year, Tuomas Sandholm and Noam Brown from Carnegie Mellon University developed a poker player called Libratus, which beat four of the world champion poker players, and not by just a little bit. They played 120,000 hands of poker, and Libratus ended up with $1.77 million in poker chips. This is a big deal, because it signals the ability to deal with incomplete information and to deal with situations that require bluffing and an opponent that generates misinformation. That is a really important set of skills. It will lend itself to negotiation, to strategy development, and perhaps even to policy analysis.
MR. AUSTIN: For decades there has been a cycle of hype and rapid progress and then you have an AI winter and it goes away. Does everyone agree it’s different this time?
MR. NG: Modern industries go through winter, winter, winter, and then eternal spring. I do think we’re in the eternal spring phase of AI, because unlike the earlier waves of maybe overhype, today AI is creating tremendous value for firms like Baidu and Google.
This creates a very clear revenue stream with which to keep investing in and improving AI technology.
JOHN BUSSEY: How might corporations that maybe haven’t thought much about AI use it to augment their strategy right now?
MR. NG: Right now, AI technology is this magical thing, right? It’s useful for so many different things. But the reality is, AI technology needs a lot of customization for your business context.
So I recommend that business leaders hire a senior AI leader—a chief AI officer or a VP—to sort this out for them.
Recruiting AI talent is so difficult that having a centralized AI function would be the best way to have consistent hiring and promotion and management standards for an AI team. This team can then work cross-functionally to figure out how to fit these technologies into your business.
MR. JACOBSTEIN: I have a different perspective on this. I believe in the power of small, interdisciplinary teams that have support high up in the corporation. It’s very important to match the speed of the technology with the nimbleness of the teams. And having a centralized AI guru at the top, where everybody has to ask questions of that person, is unlikely to be as fast and effective as having a decentralized organization with powerful teams, with real talent.
MR. AUSTIN: Do you think almost any job can be automated? Early on, we were talking about manufacturing jobs, blue-collar jobs, truck drivers. Now we’re talking about white-collar jobs.
MR. NG: Things may change in the future, but one rule of thumb today is that almost anything that a typical person can do with less than one second of mental thought we can either now or in the very near future automate with AI.
This is a far cry from all work. But there are a lot of jobs that can be accomplished by stringing together many one-second tasks.
Consider a security guard monitoring security footage. They have a pretty complex job. But the job maybe can be broken down into a lot of smaller tasks, which involve one second of cognitive thinking. So a lot of the art and skill in figuring out where to insert AI is to recognize the business opportunities where you have a complex system but a lot of these one-second tasks that you might be able to string together automatically.
MR. JACOBSTEIN: I think people are going to be surprised at how fast machine learning is going to displace routine jobs.
MR. AUSTIN: How fast are we talking?
MR. JACOBSTEIN: We’re talking about a transition that’s going to occur over the next 10 to 15 years that is really significant.
For that reason, we need to invest heavily in free education and explore various ways to provide a basic income [to those who are displaced].
MR. NG: Just as AI will destroy jobs, it will create new jobs that we can’t yet imagine. The challenge is the skills mismatch.
MR. JACOBSTEIN: The good news is that AI and robotics are going to generate massive amounts of new wealth. Our responsibility is to make sure that in addition to having our companies be successful, people who get displaced have a reasonable quality of life. So yes, we need to make education affordable because there will be new jobs.
But the real question is, “What’s the ratio of jobs destroyed to new jobs?” I think at least in the short term, that could be an unfavorable ratio.
For the past year, we as a society have been worried sick about artificial intelligence eating the jobs of 3 million truck drivers. Turns out that a more imminently endangered species are the Wall Street traders and hedge fund managers who can afford to buy Lamborghinis and hire Elton John to play their Hamptons house parties.
So maybe “hooray for AI” on this one?
Financial giants such as Goldman Sachs and many of the biggest hedge funds are all switching on AI-driven systems that can foresee market trends and make trades better than humans. It’s been happening, drip by drip, for years, but a torrent of AI is about to wash through the industry, says Mark Minevich, a New York-based investor in AI and senior adviser to the U.S. Council on
Competitiveness. High-earning traders are going to get unceremoniously dumped like workers at a closing factory.
“It will really hit at the soul of Wall Street,” Minevich tells me. “It will transform New York.”
Software is Always Learning
Some of these AI trading systems are being built by startups such as Sentient in San Francisco and Aidyia in Hong Kong. In 2014, Goldman Sachs invested in and began installing an AI- driven trading platform called Kensho. Walnut Algorithms, a startup hedge fund, was designed from the beginning to work on AI. Infamously weird hedge fund company Bridgewater Associates hired its own team to build an AI system that could practically run the operation on its own. Bridgewater’s effort is headed by David Ferrucci, who previously led IBM’s development of the Watson computer that won on Jeopardy!
AI trading software can suck up enormous amounts of data to learn about the world and then make predictions about stocks, bonds, commodities and other financial instruments. The machines can ingest books, tweets, news reports, financial data, earnings numbers, international monetary policy, even Saturday Night Live sketches—anything that might help the software understand global trends. The AI can keep watching this information all the time, never tiring, always learning and perfecting its predictions.
A report from Eurekahedge monitored 23 hedge funds utilizing AI and found they outperformed funds relying on people. Quants, the Ph.D. mathematicians who design fancy statistical models,
have been the darlings of hedge funds for the past decade, yet they rely on crunching historical data to create a model that can anticipate market trends. AI can do that too, but AI can then watch up-to-the-instant data and learn from it to continually improve its model. In that way, quant models are like a static medical textbook, while AI learning machines are like a practicing doctor who keeps up with the latest research. Which is going to lead to a better diagnosis? “Trading models built using back-tests on historical data have often failed to deliver good returns in real time,” says the Eurekahedge report.
Human traders and hedge fund managers don’t stand a chance, in large part because they’re human. “Humans have biases and sensitivities, conscious and unconscious," says Babak Hodjat, co-founder of Sentient and a computer scientist who played a role in Apple’s development of Siri. "It's well-documented we humans make mistakes. For me, it's scarier to be relying on those human-based intuitions and justifications than relying on purely what the data and statistics are telling you."
So what’s going to happen to the finance people who find themselves standing in front of the oncoming AI bus? Well, average compensation for staff in sales, trading and research at the 12 largest investment banks is $500,000, according to business intelligence company Coalition
Development. Many traders earn in the millions. In 2015, five hedge fund managers made $1 billion or more, according to an industry survey. If you think Carl’s Jr. is motivated to replace $8-an-hour fast-food workers with robots, imagine the motivation to dump million-dollar-a-year ($500 an hour!) traders.
What Happens to Traders?
Goldman Sachs shows just how devastating automation can be to traders. In 2000, its U.S. cash equities trading desk in New York employed 600 traders. Today, that operation has two equity traders, with machines doing the rest. And this is before the full brunt of AI has come into play at Goldman. “In 10 years, Goldman Sachs will be significantly smaller by head count than it is today,” Daniel Nadler, CEO of Kensho, told The New York Times. Expect the same to happen on every trading floor at every major financial company.
Much of America is not going to weep for the types of people depicted in The Wolf of Wall Street, yet this new AI reality could be devastating in many ways. Imagine the impact on high- end real estate in New York. Think of the “For Sale” signs on summer beach homes in Southampton. How will luxury retailers survive the likely dip in sales of $2,000 suits and $5,900-per-pound white truffles? Maybe Donald Trump will be driven to demand that somebody bring back traders’ jobs, thinking they’ve moved to Mexico.
Minevich, though, sees a net positive if AI drives brilliant people out of finance and into, well, almost anything else.
As the surest, fastest path to million-dollar paydays, Wall Street trading and hedge fund managing have long soaked up a large chunk of America’s best and brightest. About one-third of graduates from the top 10 business schools go into finance. Only a tiny sliver, usually around 5 percent, go into health care. An even smaller percentage go into energy or manufacturing businesses, and you can count on two hands the number who take jobs at nonprofits each year.
Most of the rest of society looks at that and sees selfishness. Yeah, sure, we need liquid markets and financial instruments and all that. But if we’re going to pay a group of people so much money, maybe we’d be better off if they were inventing electric cars that go 1,000 miles on a charge, or healthy vegetarian kielbasa, or babies who don’t cry on airplanes. Just do something that brings tangible benefits to the masses.
“Some of these smart people will move into tech startups, or will help develop more AI platforms, or autonomous cars, or energy technology,” Minevich says. That could be really helpful right now, since the tech industry is always fretting that it doesn’t have enough highly skilled pros and might be facing a geek drought in the age of Trump travel bans. If the MBA elite leave Wall Street but stay in New York, Minevich adds, “New York might compete with Silicon Valley in tech.”
As math Ph.D.’s no longer find that hedge fund recruiters are salivating over them, they might leap into efforts to model climate change or the behavior of cancer cells in the body. The National Security Agency’s website says it “is actively seeking mathematicians to work on some of our hardest signals intelligence and information security problems.” Math whizzes could help catch terrorists! Or liberals!
The pay for a mathematician at the National Security Agency is around $100,000. Compared with a hedge fund salary, that would be a major lifestyle downgrade. But at least the traders and quants will have options, which is more than we can say for truck drivers and other workers threatened by AI.
There’s one other benefit to AI machines taking over finance. Ben Goertzel, chief scientist at Aidyia, says his machine will never need human intervention. “If we all die, it would keep trading,” he once said.
So if Trump pulls out the nuclear codes and pushes the button, at least some people will still get a good return on their 401(k)s.