WHAT IF one day, maybe soon, robots and computers become silicon Einsteins that are smarter than humans? Would they rule us and harm us?Two of the world’s cleverest entrepreneurs, along with a brilliant physicist and leading journalist, disagree sharply on the answer.Elon Musk, founder of Tesla and SpaceX, recently told the US National Governors Association, “I keep sounding the alarm bell, but until people see robots going down the street killing people they don’t know how to react.” Musk wants strong government regulation of this technology. Earlier, in 2014, he called artificial intelligence (AI) “our biggest existential threat” and said AI could continually improve itself until it viewed humans as obsolete.That same year, British cosmologist Stephen Hawking told the BBC, “The development of full artificial intelligence could spell the end of the human race... [The silicon Einstein] would redesign itself at an ever- increasing rate… Humans, who are limited by slow biological evolution, couldn’t compete and would be superseded.”In response, Facebook founder Mark Zuckerberg called Musk a “naysayer” and called his doomsday fears “pretty irresponsible.”He added, “In the next five to 10 years, AI is going to deliver so many improvements in the quality of our lives.”Musk retorted that “Mark’s understanding of the subject is limited.” New York Times columnist Thomas Friedman explains, “We’re moving into a world where machines and software can analyze (see patterns that were always hidden before); optimize (tell a plane which altitude to fly each mile to get the best fuel efficiency); prophesize (tell you when your elevator will break and fix it before it does); customize (tailor any product or service for you alone); and digitize and automate just about any job.”It is transforming every industry. Who is right? Musk and Hawking, or Zuckerberg and Friedman? For examples of AI that are boons to humanity, take, for instance, Alexa and Watson. Alexa is an intelligent personal assistant developed by Amazon. Watson is a powerful IBM computer and related software able to answer tough medical questions framed in ordinary language.
ALEXA WAS introduced in 2014. She talks, answers questions, plays music, makes to-do lists, sets alarms, does streaming podcasts, plays audiobooks and provides weather, traffic and other real-time information, such as news. Alexa also can control several smart devices using itself as a home-automation system.Most devices allow Alexa users to activate the device using a wake-word. Alexa speaks English and German, but more languages will be added soon.Writing in The New York Times, Penelope Green says, “Since her introduction in November 2014, Alexa has neither devolved into the malevolent intelligence predicted by science-fiction writer Arthur C. Clarke nor ascended to the metaphysical eroticism promised by Spike Jonze (by way of Scarlet Johansson, who plays Samantha, a computer operating system) [ in the movie “Her”]. Instead, she has assimilated as a kind of ideal roommate with none of the challenges of an actual human being. This year, 25 million Americans will use an Alexa device at least once a month….”Is Alexa intelligent? With her voice disguised, Alexa can easily answer questions in ways that could be attributed to a human being. This is the so-called “Turing test” for computer intelligence – you could not distinguish Alexa’s response from that of a human being. While Alexa helps plan our day, Watson tries to cure cancer.On May 11, 1997, history was made; IBM developed a super computer it called Deep Blue that knew how to play chess. A match was arranged with the reigning world chess champion, Gary Kasparov. Deep Blue won.Was Deep Blue intelligent? Yes and no.No, because it was simply able to calculate an enormous number of possible chess moves in a fraction of a second. Speed is not intelligence. But, yes, because it was able to analyze these chess moves and pick the best one sufficiently well to beat Kasparov.That dramatic chess-playing computer sounds a bit frivolous – but it generated enormous interest and massive research and development. In 2010, IBM continued with its viral high-visibility demonstrations of Big Blue technology. An IBM computer played the TV quiz game Jeopardy and defeated two of its greatest human champions, Brad Rutter and Ken Jennings, and it wasn’t even close! Now, IBM has developed a much more useful computer/ software system ‒ known as Watson ‒ named not after Sherlock Holmes’s assistant, Dr. Watson, but after IBM pioneer Thomas J. Watson.Watson has been taught to understand cancer. It reads thousands of research papers published monthly and learns their content. It then answers questions, posed in ordinary English, from cancer researchers, often coming up with treatments that even highly experienced and intelligent oncologists did not think of. Watson has already prolonged human lives, and perhaps even saved some.When the Internet of Things arrives and billions of devices are networked, some experts say knowledge will double every 12 hours. Human intelligence and learning will not be able to keep up. Even today, no oncologist can read even a tiny fraction of the published cancer research papers. So, we will need the help of intelligent computers.The issue of whether AI is benevolent or malevolent is highly relevant for Israel’s hi-tech industry because a number of homegrown start-ups are among the world’s leading pioneers of applied artificial intelligence; they include RealFace, Zebra Medical Vision, Voyager Labs and Waycare. The IDF, too, is an AI pioneer.But, before I describe them, it is first important to understand what exactly is meant by the term artificial intelligence and related concepts because they are generally poorly understood by us non-experts.Artificial Intelligence: The term is very misleading. Intelligence is the ability to solve problems and to learn. Learning and problem solving are related. The more problems you solve, the more you learn. And the more you learn, the better you get at solving problems.Human beings have intelligence. So do animals. And so do some machines. If a human being, a chimpanzee and a computer can learn and solve problems, then they all have intelligence. There is nothing artificial about the silicon variety.IN FEBRUARY, Apple acquired RealFace, an Israeli AI facial-recognition company founded by Aviv Madar and Adi Barzilai. RealFace technology provides “frictionless face recognition” directly on a relatively low-powered device such as an iPhone and aims to “offer customers a smart biometric log-in solution” for mobile phones.A small IDF Military Intelligence unit with 20 career officers is leveraging AI for crucial battlefield operations. According to the daily Haaretz, the unit is using AI to predict where and when Hamas rocket launchers in Gaza will be set up and to identify and name suspicious objects in video-camera footage.Machine Learning: Machine learning is a type of AI that comprises a set of computer algorithms that get better and better with experience, i.e. they know how to learn from experience and make predictions that get more and more accurate. Machine learning has evolved from pattern recognition, which focuses on the recognition of patterns and regularities in data.A machine-learning computer learns from experience if it undertakes some task and gets better at it with experience. One type of machine learning is unsupervised learning ‒ the computer is asked, “Computer, what do these data mean? and is then told, ‘You’re on your own.’”Take, for example, Waycare, a start-up co-founded by Idan Hahn, Shai Suzan and Noam Maital. (Full disclosure: Noam is my son.) Waycare is a prediction platform that uses machine learning and deep learning to predict traffic accidents or crashes about two hours before they occur and gives city officials, especially traffic managers, a forecast tool for up to 24 hours in advance, or a tool to control traffic signals to optimize traffic flow. The goal is to change the way cities manage accidents from a reactive perspective to a proactive perspective.Today, cities send police and emergency medical responders when accidents are reported. Waycare’s technology can help prevent such accidents and position responders in advance near accident-prone stretches of road, greatly shortening the crucial response time. In September, a full-scale trial will get under way in Las Vegas.One day, the roads will be filled with self-driven (autonomous) vehicles. Each will generate four terabytes of data every eight hours. Only machine learning will be able to process such mountains of data to smooth our way home and to work.Data Mining: Data mining and machine learning are not the same thing. Machine learning focuses on prediction ‒ what can you, computer, predict from the data we give you? Data mining focuses on finding previously unknown properties of the data, e.g. finding that stock market prices in the past were correlated with the 11-year sunspot cycle. Data mining focuses on past knowledge; machine learning seeks to create new knowledge.Zebra Medical Vision, founded in 2014 by Eyal Toledano, Eyal Gura and Elad Benjamin and based on Kibbutz Shefayim, teaches computers to read medical images by combining a large imaging database with deep learning techniques to build algorithms to automatically detect and diagnose medical conditions.Deep Learning and Neural Networks: Deep learning technology is closely related to neural networks.The human brain is a most incredible organ. There are about 86 billion neurons or brain cells, connected by synapses – tiny “cables” that allow one brain cell to communicate with another by transmitting a very small electrical signal. The linkages of many brain cells are called neural networks.The brain itself is a complex neural network comprised of a large number of subnetworks. The more such connections, the better our brain is able to learn, think and create. Man-made neural networks have far fewer neurons than human brains ‒ perhaps, the size of a flatworm’s brain ‒ but, nonetheless, can do amazing things.Deep learning is “deep” because the learning takes place through an “artificial” or man-made neural network that is several layers deep. Just as the brain works by sending electrical signals from one neuron to another many many times ‒ passing through a variety of locations in the brain and through a variety of subnetworks ‒ so does deep learning “learn” by employing several layers of neural networks.UNSUPERVISED LEARNING starts at a low-level neural network, is passed on to a higher level one, moving all the way up until the learning is satisfactory or complete.For example, Voyager Labs, founded in 2012 by Avi Korenblum, uses deep learning to understand individual behaviors, interests and intent based on public data widely available to all “but not understandable by all.” Voyager products can measure credit worthiness or offer product recommendations.In science-fiction writer Arthur C. Clarke’s movie “2001: A Space Odyssey,” HAL 9000 (a heuristically programmed algorithmic computer) uses AI to control the Discovery One spacecraft. Hal becomes malevolent and kills the whole crew except Dave Bowman, who manages to shut Hal down.More than 70 years ago, science-fiction writer Isaac Asimov, perhaps the greatest of them all, wrote the short story “Runaround,” which defined his three laws of robotics: A robot may not harm a human; a robot must obey human orders; and a robot must protect itself unless doing so violates laws one and two. Prophetically, he later added a fourth law: A robot may not, by inaction, allow humanity to come to harm.Will the silicon Einsteins that Israeli entrepreneurs are helping to develop enrich and prolong our lives and change the world? Will they obey Asimov’s rules? Or will they one day tyrannize and enslave humanity? AI is no different from history’s other life-transforming technologies, e.g. nuclear energy or genetic engineering. All are capable of benevolence or destruction. It is up to us which result we choose. The writer is senior research fellow at the S. Neaman Institute, Technion and blogs at www.timnovate.wordpress.com