Artificial intelligence—a new machine age
When computers mimic the capabilities of the human brain, that is artificial intelligence (AI). From the outside, AI looks like computers that have independent thoughts. Have no fear, however. The gears of their machine “brains” may be turning, but, for right now, they’re not really thinking—at least not the way that human beings think.
Popular culture has held artificial intelligence in great esteem for decades. The excitement around conscious robots has grown as computers have become ingrained in every aspect of our lives.
You don’t have to look long for examples of artificial intelligence. And, the applications are diverse:
- Self-driving car: A Wired® magazine reporter got behind the wheel of—but did not drive—a “piloted” Audi® A7 for a 500-mile excursion. “It’s so sophisticated that I never felt anything unusual,” the reporter wrote.
- Robot butler: The Aloft® hotel in Cupertino, Calif., dispatches cute, 3-foot-tall rolling robots with toiletries and food. The robot asks guests to rate the delivery service on a touchscreen before saying goodbye.
- Document examiners: E-discovery software was used to sift through 1.5 million documents in about a month for less than $100,000. This task would have been performed by paralegals and lawyers for a significantly higher cost.
- Game show champ: The IBM® Watson™ computer defeated two human “Jeopardy!®” champions. IBM is developing versions of Watson for health care and retail applications.
Yet, for all its promise, artificial intelligence seems to live under a dark cloud. Elon Musk, the tech industry superstar who runs an electric car company and a space exploration company, said AI is “summoning the demon.” And, movies about the dangers of AI are too numerous to mention.
Artificial intelligence is opening exciting doors in our personal and professional lives, but the technology is not without risk—real and perceived. Taking advantage of the opportunities offered by artificial intelligence calls for understanding the basics of AI, examining applications of the technology and exploring its pros and cons.
A crash course in artificial intelligence
Artificial intelligence exists because of deep learning (or machine learning). And, deep learning starts with neural networks. Neural networks simulate the web of neurons in the human brain. Data make neural networks smarter. The idea behind training the neural network is this: “Feed enough photos of a dog into these neural nets, and they can learn to recognize a dog. Feed them enough human dialogue, and they may learn to carry on a conversation.” Training neural networks is the engine of deep learning.
As the neural net is trained, the computer has a job to do: “Come up with a statistical rule that correlates inputs with the correct outputs.” This rule is a deep learning algorithm, and its accuracy improves with additional data and time—just like the performance of the human brain.
Deep learning is edging machines toward almost humanlike understanding:
“Machine learning is a way of getting computers to know things when they see them by producing for themselves the rules their programmers cannot specify. The machines do this with heavy-duty statistical analysis of lots and lots of data.”
Programming deep learning algorithms does not require subject matter expertise. For example, a team in Switzerland developed a deep learning algorithm that’s been used to read Chinese at the native speaker level. No one on the team can speak or understand Chinese. An algorithm used at the world-renowned European Organization for Nuclear Research (CERN) physics lab helped spot subatomic particles better than software written by physicists. Programmers set the wheels of deep learning in motion, and computers make continuous improvement.
The human brain solves complex problems by having sets of rules, and it’s easy to translate rules into programming. That’s why artificial intelligence excels at complex tasks with explicit rules. To compare, the human brain is naturally good at tasks with implicit rules—facial identification, recognizing speech, picking out objects in pictures. AI can complete these simple tasks with a lot of data and training.
Today’s artificial intelligence landscape
Business spending on artificial intelligence is exploding. In 2010, venture capital (VC) spending on AI was less than $20 million. In 2014, it was close to $300 million. Figure 1 below shows the rapid rise in AI dollars:
Figure 1: Total venture capital money spent on AI startups
This rapid rise in VC funding indicates that industry sees AI as the future. Examples of AI investment include:
- Google®: The search engine uses RankBrain, an AI system to interpret search queries. RankBrain has a success rate of 80 percent when it comes to guessing which pages should rank on top. To compare, human engineers guessed right 70 percent of the time.
- Facebook®: The social network has an AI lab, and it uses AI to personalize homepages.,
- Baidu®: The leading Chinese search engine hired away Andrew Ng, the data scientist who helped lead deep learning at Google. Deep learning could improve Baidu’s speech recognition technology, which is important to illiterate smartphone users. Ten percent of Baidu’s searches are by voice, and that number could jump to 50 percent by 2020.
- Forbes®: The financial news outlet trusts technology by robo-writing firm Narrative Science® to automate the writing of basic financial stories.
All these examples show that industry-leading organizations are putting money into AI. Most notable, though, is that they trust AI with their profit centers. Artificial intelligence and deep learning are essential to product development and customer satisfaction.
The biggest buzz in AI today surrounds quantum computing. Google’s D-Wave 2X quantum computer has grabbed attention with tests showing that it’s 100 million times faster than any standard computer. Harmut Neven, engineering director of the quantum computer experiments, said the fast performance “will carry over to commercially relevant problems as they occur in tasks relevant to machine intelligence.”
What does this mean to someone who’s not a computer expert? In short, quantum computers are very good at optimization problems, which happen when you try to reach a solution by taking thousands of variables into consideration. For instance, if you were planning a trip to Europe, you could tell a quantum computer which cities you’d like to visit and how much you want to spend. The computer would give you an itinerary based on those elements and even more factors, including:,
- Baggage fees
- And, many more
Optimized solutions are derived from volumes of data—volumes of data that train deep learning algorithms. The future of artificial intelligence may be tied to quantum computing and optimization:
“Imagine NASA being able to use quantum computers to optimize the flight trajectories of interstellar space missions, FedEx® being able to optimize its delivery fleet of trucks and planes, an airport being able to optimize its air traffic control grid, the military being able to crack any encryption code, or a Big Pharma company being able to optimize its search for a breakthrough new drug.”
Part of the reason quantum computers are so powerful is that they work differently from standard computers. A standard computer reads everything in 1s and 0s. A quantum computer, to compare, reads everything in 1s and 0s, and sometimes a 1 and a 0 can exist at the same time.
Google’s D-Wave 2X isn’t the only quantum computer on the block. The Yale Quantum Institute also is exploring quantum computers. As commercial interest in artificial intelligence has grown, so has the market for AI products and services.
Page 1 of 3