Chapter 7: Evidence of Automation

We understand what exponential growth means. We have seen how information technology has grown over the last 150 years. Let us see how far has that brought us.

I started gathering the evidence for this chapter as soon as I decided to write the book in October 2011. Since then, I have collected more than 300 articles, all from reputable and reliable sources. These stories cover machines that act like us, computers that “think” better then us, and robots that perform unimaginably complex tasks. Every day I opened my news feed to find something new and then add it to my list. At a certain point I realised I had to stop. I knew there could never be an end to this trend, but I did not expect it to grow so quickly. Once again, I underestimated the power of the exponential function. As the list started to grow out of proportion, I decided I would freeze it, finish the book, and publish, or else I would never finish it. To offer readers a current resource, I will continue to post updates on the website http://robotswillstealyourjob.com . In this book, rather than a sterile and long list of technologies, I picked only a few that I think are most relevant for the sake of the argument.

1.1 Automated Shopping

You might not think of them as such, but vending machines are actually a primitive type of robot. Their function is very simple. They keep an inventory, have an electronic display, accept money, and provide you with the item you purchased. It is a 30-year-old technology and it has not progressed much since then. Or, has it? In Europe and the US we do not think much of vending machines, but that is just because we have not taken them seriously. In Japan, however, where they have high population density, limited space, high cost of labour, low rates of vandalism and petty crimes, and people shop mostly by bicycle or on foot, vending machines are taken very seriously.

In Japan there are about 8.6 million vending machines, one for every 14 people, the highest number pro capita in the world.1 . These robots, known there as jido-hanbaiki (from jido, ”automatic”; hanbai, ”vending”; and ki, ”machine”), often abbreviated jihanki, are widespread and commonly used for all sorts of goods: not just newspapers, snacks, and drinks, but also books, DVDs, condoms, ice-cream, hot instant noodles, rice, magazines, glasses, boiled eggs, umbrellas, neck ties, sneakers, vegetables, iPods, live lobsters, Onsen (hot spring water), and even Buddhist prayer bead-rolls. Sure, we can laugh at it, but doesn’t it make sense? The days when you had the little shop just around the corner – with a smiling person who owned the shop, knew what they were doing, and could give you real guidance and assistance – are quickly disappearing.

Most commercial transactions of physical goods today are made at the mall and huge supermarket chains. The cashiers at these companies work part-time, as one of the multiple jobs they hold because just one job will not provide the money they need to pay their rent, medical bills, student loads, mortgages, etc. The truth is that it would make a whole lot of sense for society to have a shopping mall where most things are automated. The problem with that, of course, is that people currently working there would find themselves in deep, deep trouble.

Imagine this. You walk into a store and you have an interactive map on your cell phone showing you where all the items are. You can search for items, filter them by categories, and get information on each single product that goes far beyond nutrition facts; you can trace the production process, the companies behind it, and dynamically compare products based on your search criteria. You can also read reviews from other people about these products, just like on Amazon.com today. Before leaving with your items, you stop for a few seconds in an assessment zone that receives signals from RFID chips in the merchandise. Then you swipe your credit card, or just accept the payment request on your cell phone. The whole process, the time between when you decide to leave the store and the moment you can actually walk out, takes less than 10 seconds. No human was involved in this, no human was required. No queues, no waiting time.

Sound futuristic? Every piece of technology needed to make this happen already exists, and has existed for many years. Then why is it not in place already? Why are we not seeing this trend expanding to all retail stores? Maybe it is expensive to deploy such a system. Actually, it would be much cheaper than having to employ humans to do the job. “But you need human contact! What about the value added that only a human employee can offer?” Have you ever worked at a mall? If so, how motivated were you and how long did that last? “But you need human workers to place the products on the shelves!” Actually, even that technology is already available, though it is more recent than the others. Some warehouses are already completely automated, and require only operators to work and handle the entire task. Pallets and product move on a system of automated conveyors, cranes, and automated storage and retrieval systems coordinated by programmable logic controllers and computers running logistics automation software. Their accuracy and productivity far exceeds that produced by human labour. These machines are faster, more precise, they can lift huge weights without having to deal with back problems, they work day and night, and do not require much maintenance. Amazon.com recently made a $775 million purchase of Kiva Systems, a manufacturer of bright orange robots that scuttle around warehouses filling orders2 . CNN has a video of the system operating (see the link in the previous footnote or the book website). It is a pretty amazing sight. Hundreds of robots transporting merchandise around immense warehouses, with clockwork precision and perfect timing, as if dancing to a silent piece of music written in code of zeroes and ones. These robots are smart enough to put the items in the most convenient place and distance, based on how frequently they are needed, how heavy they are, and many other criteria. They work 24/7 and never make mistakes. The application of similar automated systems to supermarkets and shopping malls is a minor engineering issue, one that can easily be solved in a few months, if there was ever the intention to do so.

If this is all possible, why are we not seeing it?

Tesco is the third-largest retailer in the world measured by revenues (after Walmart and Carrefour) and the second-largest measured by profits (after Walmart). Tesco has a large market in South Korea (where they are branded as “Home plus”), second only to E-Mart mainly because that company has more stores. As one might expect, they wanted to increase profits. The typical approach would require them to build more stores in order to reach E-marts level of distribution in the country. They opted for a different strategy, one that uses more automation and less workers.

Picture yourself in Korea going to work. You need a few things for dinner, but don’t have much time. While waiting for the next subway train to arrive, you see the walls covered with displays that look like supermarket shelves. You use your cell phone to scan the QR code on the items you want and then check out. When you get home, you will find your groceries have been delivered to your doorstep. Quite convenient, isn’t it? The results of this experiment are in: online sales between November 2010 and January 2011 increased by 130%, with the number of registered members rising by 76%. Home plus had become the number one online store, while successfully raising the stakes in the offline market3 .

This continuing trend could potentially destabilise the economy. Consider the millions of employed who would be affected by it. If Walmart were to put this technology in place on a systemic level (automated restocking, shopping, and delivery), the consequences to those currently employed by them would be disastrous. It would be practically impossible for most of them to find another job. The average person does not realise how big Walmart really is. Today, Walmart is the largest retailer on the planet. In fact, it is much more than that: the finances, footprint, and personnel of this behemoth dwarf entire industries and countries4 . Its epic $421 billion annual revenues eclipse the GDP of more than 170 countries. Its 2.1 million employees could form the second largest standing army on the planet. Walmart’s 2010 revenues were bigger than the revenues of the largest oil companies, the largest manufacturer, and the largest pharmaceutical company in the United States. Even when combined, the revenues of Chevron, General Electric, and Pfizer still total less than Walmart’s. To put this in perspective, if Walmart were a country, its GDP would be the 25th largest economy in the world (twice the size of Ireland’s). If Walmart were to initiate an aggressive automation strategy, in just a few years it could easily run its business with less than 100,000 employees. That would leave 2 million people, mostly uneducated and unskilled workers, out of a job. Where would these people go? And what would they eat? What will happen to their families?

In the past, we have seen automation cutting the workforce, but unskilled workers all gravitated towards places like Walmart to find an easy (although very unsatisfying) job. This is one of many unspoken tragedies of the so-called modern culture. The idea that the greatest aspiration a person could have is to work some mechanical and monotonous job, so that they can pay the bills, is an insult to the dignity of every individual. Each human being, from the moment they are born, is an invaluable masterpiece, capable of greatness beyond what we can conceive today. To even consider the proposition that we should hang on to an economic system that hinders innovation and automation, in order to preserve repetitious and mindless jobs, shows the deep loss of perspective and aptitude of our out-dated institutions.

If Walmart begins automation (and I suspect they will), there would be no coming back for the shopping industry. It is an irreversible process. The replaced jobs will not come back. But having removed these jobs, what will millions of people do?

Wait before you answer, we are not quite done yet.

1.2 Automated Manufacturing

The advent of automation in the manufacturing industry is generally well-known. It has been a century since we started using machines to increase our productivity. Just think of a car factory. The assembly line developed by Ford Motor Company between 1908 and 1915 made automated assembly widespread and mass production brought unprecedented social transformations. By utilising the old Latin proverb divide et impera (divide and conquer), we were able to transform long and difficult tasks into sets of many small and simple-to-execute mechanical operations. This approach worked well with machines which, for a century, integrated with humans in a fruitful cooperation.

Robots were displacing human workers, but we always found something else to do, because of mainly two reasons:

  • There was enough time to adjust and learn new skills.
  • Some operations were too complex for machines to do, or the cost of creating a machine capable of performing such a task was too high. Why go through the trouble of programming a complex robot to do something cheap labour could accomplish more easily and at less cost?

Such was the past, but things are different now. Labour is no longer so cheap. Human development is finally occurring on a mass scale.People are (justifiably) demanding their rights. Even though there are still millions who work in conditions that we might consider slavery by today’s standards, the working conditions and standards are raising everywhere, even in relatively under-developed countries. On the other hand, however, algorithms are exponentially improving, robotics technology is developing rapidly, and machines are now becoming cheaper to build (even for complex tasks). We are already seeing the effects of this everywhere.

Foxconn is the largest maker of electronic components in the world and the largest exporter in Greater China,5 6 with an annual revenue of more than 100 billion dollars.7 They make virtually anything. If you have an iPad, an iPhone, a Kindle, a PlayStation 3 or an Xbox 360, chances are very good that Foxconn made it. Without counting national public services, Foxconn comes out as the third largest employer in the world with an impressive 1.2 million workers, right after Walmart (2.1 million).8 It has contracts with Acer, Amazon.com, Apple, Cisco, Dell, Hewlett-Packard, Intel, Microsoft, Motorola, Nintendo, Nokia, Samsung, Sony, Toshiba, and just about any major tech company you can think of. Foxconn is not a company: it is an immense monster, an electronics supergiant that is singlehandedly responsible for nearly half of all such technological production in the world.9

If they were to displace their 1.2 million workers, things would turn ugly for many people. As it happens, as recently as last year (2011), Foxconn announced that they intended to deploy an army of robots in order to “replace some of its workers with 1 million robots in three years to cut rising labor expenses and improve efficiency.” – said Terry Gou, founder and chairman of the company.10 It still remains unclear if they are really going through with the plan, and how many workers would be displaced by this initiative, but it appears that they have already launched and built a Research and Development facility and a factory in Taiwan to build their own robots; and have begun to hire some 2,000 engineers to drive the project forward.11 It appears that Foxconn is committed to the automation of their business, and it should come as no surprise. Why wouldn’t they? Robots are cheaper and more reliable than human workers, they do not ask for vacation, they do not commit suicide, they do not protest for more rights, and they can ensure the company’s profits – which is what matters most for a multinational corporation and its stakeholders.

Rumors and stories surrounding Foxconn’s operations began to spread after a wave of suicides was reported by the news in the Western world. After fourteen workers were found dead in 2010, some twenty Chinese universities compiled a report in which they described Foxconn factories as labour camps and detailed widespread worker abuse and illegal overtime.12 Stories of overcrowding, tiny living accommodations, impossibly long and exhausting work hours, and security guards beating workers to death are just hints of what happens in those hellholes; what manages to overcome the great firewall of censorship of China to reach our digital shores.13 After protests began to kindle in the US and in Europe, demanding better working conditions, the morbid response from Foxconn executives was that they would install suicide-prevention nets at some facilities to catch the people who tried to commit suicide by jumping off the building (I am not joking), and they promised to offer higher wages at its Shenzhen production bases. But they also did something else. Workers are now forced to sign a legally binding document guaranteeing that they and their descendants would not sue the company as a result of unexpected death, self-injury, or suicide.14

The saddest part of this story is not the that workers at Foxconn live in horrifying conditions. What is truly astonishing is that Foxconn actually provides higher wages, better working conditions, and has a lower suicide rate than the average Chinese company.15 Foxconn is merely the story that made it into the news and we suddenly became all outraged by it. But there is nothing to be surprised about: this is the very nature of current socio-economic system, efficiency and, consequently, profits are more important than human lives.

Foxconn is not the only company moving in the direction of automation. Canon announced in June 2012 that some of its camera factories will phase out human workers in an effort to reduce costs. We can expect robots to be making the next generation of cameras, possibly as soon as 2015. Of course, the company’s spokesman Jun Misumi was quick at dismissing the idea that this move would mean layoffs at Canon when he told the Associated Press, “When machines become more sophisticated, human beings can be transferred to do new kinds of work“.16 These are nice words, but I doubt they will hold true. Assembly line workers have been performing the same mindless, repetitive, mechanical jobs for years. Before they started working at a factory, they were masterpieces of evolution and natural selection, individuals with imagination, dreams, and aspirations. They had endless possibilities. They could have become artists, scientists, and musicians. They could have been the drivers of new amazing discoveries that pushed humanity forward. After a few years in a factory they each were just another pair of hands in an endless sea of moving parts, their dreams were crushed, their hopes and aspirations reduced merely to bringing home just enough to keep their heads above water for another month. I doubt these people will all suddenly become engineers, industrial designers, sales managers, and computer scientists – assuming that a proportionally larger number of those jobs will be needed at Canon by 2015 (they will not).

Foxconnn and Canon are only two of numerous examples. China is increasingly replacing its workers with robots 17 and now even major newspapers are realising this. Just a few days ago (at the time of this writing), The New York Times came out with a 6-page piece titled “The Machines Are Taking Over”18 and The Wall Street Journal says “Why Software Is Eating The World”.19 I suspect these types of articles will only increase in the near future.

The trend is clear. Companies in the manufacturing sector are automating and the typical statement that “people will find something else to do” is simply a cop-out that does not look at the reality of the situation – that change is happening too fast and that most workers who will be replaced by machines will not have the time to learn new skills. Assuming, of course, that we could somehow find a number of new jobs equivalent to the number of displaced workers – I very much doubt we will (more on this in Chapter 9).

1.3 3D Printing

You are in your house having a party with some friends. As it happens, one of them drinks a little bit too much and drops a glass on the floor. Typically you would have to go out and buy a new one, or get online and order it. But, you could also go the computer, download the CAD file of the glass, press print, and watch your 3D printer as it makes a perfect replica of the glass to replace the one your friend broke. Pretty neat, but not really a game changer.

Now imagine you are Captain of a container ship. You left from China a few days ago on your way to San Francisco and now you are in the middle of the Pacific Ocean. Suddenly the ship stops and the Chief Engineer comes to the bridge to tell you that a part of the engine just broke. He does not have a spare part and has no way of making a replacement. You realise you are stranded. All you can do is call for help, wait, miss the deadline, and lose a lot of money. Not a pretty situation. Or, you could have a 3D printer. Select the file, press print, fix the engine, and be on your way in less than an hour. That is pretty neat.

It is like the replicator in Star Trek20 . “Tea. Earl Grey. Hot.” Many fans of The Next Generation will recognise these words. Just say the word and anything you want will appear right in front of your eyes. How far are we from this fantastic technology?


PIC

Figure 1.1: The replicator in Star Trek creating a coffee mug.

Today 3D printing is a multi-billion dollar industry, and it is growing exponentially21 . There are many types of 3D printer, from DIY Open Source models to sophisticated commercial products, spanning from a few hundred to many thousands of dollars. The idea behind it is simple. Just like regular inkjet or laser printers, they start from a file on your computer and then manipulate matter to create what you want. The only difference is that they can print in three dimensions instead of two, and they can use many different materials. 3D printers are already used for rapid prototyping, rapid manufacturing, and many DIY enthusiasts and hackers use them at home for fun. Although these machines are not quite ready to replace all commercial production, they surely are on their way. The hugely successful Open Source project RepRap gave rise to a plethora of successors, thanks to its openness and incredible community of people around it. Just to name of few of the available 3D printers under 1,000, we have MakerBot Thing-O-Matic, The Replicator, Ultimaker, Shapercube, Mosaic, Prusa, Huxley, Printrbot. They all came into existence in just a couple of years, and if you buy it in kit form and assemble it yourself, you can get one for less than 300.


PIC

Figure 1.2: The “Replicator”, an inexpensive 3D printer that prints object in colours.

Printers in the lower price range are still very limited, both in terms of resolution (you can see the imperfections) and the materials they can use (mostly plastics). However,commercial printers are different. At the time of this writing, the most sophisticated machine can print with an accuracy of 16 micrometres22 . That’s 0.016 millimetres! To put things in perspective, the resolution limit of the human eye is about 100 micrometres, and the iPhone 4’s ‘Retina display’ pixels are 78 micrometres in width23 . These machines can print multiple materials, such as ABS plastic, PLA, polyamide (nylon), glass filled polyamide, stereolithography materials (epoxy resins), silver, titanium, wax, polystyrene, ceramics, stainless steel, titanium, photopolymers, polycarbonate, aluminium and various alloys including cobalt chrome.24 You can print in colour and even create structures that are more intricate than any other manufacturing technology – or, in fact, are impossible to build in any other way.25 You can create parts with moving components, hinges, and even parts within parts.

3D printers are not just used as an alternative to standard manufacturing. People have printed really cool-looking personalised prosthetic limbs,26 bone-like material,27 and even human organs.28 29


PIC

Figure 1.3: A 3D printer-created lower jaw that has been fitted to an 83-year-old woman’s face in what doctors say is the first operation of its kind.

A very inspiring example of how 3D printers can be used for the betterment of humankind comes from Scott Summit and his team composed of Industrial Designers and Orthopaedic Surgeons, whose mission is to bring more humanity to people who have congenital or traumatic limb loss. In their words: “Each of our bodies is unique, as are our tastes and styles. Humans are anything but one-size-fits-all, and we want to recognise that fact. We achieve this by creating products that allow our clients to personalise their prosthetic legs. Our hope is to enable our clients to emotionally connect with their prosthetic limbs, and wear them confidently as a form of personal expression.”30 For people who have lost a leg, life can be very difficult. So, instead of hiding their defect and feeling ashamed of it, they can show it with pride, reclaiming that lost connection with their body.


PIC

Figure 1.4: Beautiful pictures of 3D printed prostheses. Courtesy of Bespoke Innovations, Inc.

I expect we will soon see a rapid increase in the quality of these machines, with the costs dropping so significantly that they will become an everyday commodity, much like a microwave oven can be found in most houses. Marketplaces like iTunes, Android, and Amazon.com will follow, along with their ’pirate’ and Open Source counterparts. In fact, the Open Source community is already leading the way (as always). Thingiverse has thousands of free designs that people can download, print, or improve upon,31 and The Pirate Bay recently announced a new section called “Physible,” CAD designs of physical objects, legal or not32 . In a few years, most of us will all have a micrometre-precision 3D printer that prints multiple materials and colours in our house. Designs will be extremely cheap, or they will cost nothing at all.

Today 3D printing is little more than a hobby, but it will probably soon become a game changer for entire industries. Another advantage of 3D printing is that instead of conforming to sizes and shapes defined by the economies of scale, the object can adapt to you, instead, moving from an economy of mass production to an economy of mass personalisation. How many jobs today rely on manufacturing? We’ll probably see them disappear, too.

1.4 Automated Construction

Typically, it can take from 6 weeks to 6 months to build a 2-storey house in the US or Canada, mostly because dozens of humans do all the work. However, we have newer and smarter ways of building houses, which some are beginning to use. In China, we can construct a 30-storey skyscraper with all modern comforts in 15 days. That’s 2 storeys per day, non-stop. The building is made from prefabricated parts and can withstand earthquakes of magnitude 9. It has excellent insulation systems, is five times more efficient than regular hotels, and has smart systems for air circulation and quality control33 . The implications of this are significant: we have designed a system that will let you build anywhere, to construction tolerances of +/- 0.2 mm, in just a few days34 .

This is what we can do today. Let us have a look at tomorrow, shall we?

Contour crafting is a construction process that uses a computer-controlled crane or gantry to construct buildings rapidly and efficiently without manual labour. It is possible that within a decade this technology will advance so much that we will be able to upload the design specification to our computer, press print and watch massive robots spit out a concrete house in less than a day. No humans required, except for a few supervisors and designers. You might be thinking this is like a huge 3D printer! And you would be right. The idea is the same, just the scale and the materials differ.

Contour crafting is now under development by Behrokh Khoshnevis of the University of Southern California’s Information Sciences Institute. It was originally conceived as a method to construct moulds for industrial parts, but Khoshnevis decided to adapt the technology for rapid home construction as a way to rebuild after natural disasters like the devastating earthquakes that have plagued his native Iran.35 Khoshnevis claims that his system could build a complete home in a single day, and its electrically powered crane would produce very little construction material waste. This is particularly interesting because today a standard home construction project creates 3 to 7 tonnes of waste, as well as exhaust fumes from machinery and vehicles,36 not to mention the thousands of deaths each year which result from workplace accidents.37 Contour crafting could reduce costs, lessen our environmental impact, and save materials and lives. Of course, many jobs would disappear, too.

Some industries and institutions have already shown interest in this technology. Caterpillar, Inc. has provided funding for the Viterbi project since the summer of 2008,38 NASA is evaluating Contour Crafting for its application in the construction of bases on Mars and the Moon,39 and Singularity University graduate students established the ACASA project with Khoshnevis as the CTO to commercialise Contour Crafting.40

1.5 Automated Journalism

You might think that writing is one of those things that machines will never do. Sure you can program them to generate text, but it will sound sterile and fake. It would have no soul. You would be able to spot it in a second, right? Right?

Let us see how well you do. Below are the opening lines of three story pieces written about a baseball game. Can you tell which were written by flesh-and-blood humans, and which (if any) were written by a computer?

a)
The University of Michigan baseball team used a four-run fifth inning to salvage the final game in its three-game weekend series with Iowa, winning 7-5 on Saturday afternoon (April 24) at the Wilpon Baseball Complex, home of historic Ray Fisher Stadium.
b)
Michigan held off Iowa for a 7-5 win on Saturday. The Hawkeyes (16-21) were unable to overcome a four-run sixth inning deficit. The Hawkeyes clawed back in the eighth inning, putting up one run.
c)
The Iowa baseball team dropped the finale of a three-game series, 7-5, to Michigan Saturday afternoon. Despite the loss, Iowa won the series having picked up two wins in the twinbill at Ray Fisher Stadium Friday.

Take a moment and try to guess. They all look pretty similar, but which one is the product of a lifeless machine? All of them? None? It is time for the moment of truth. If you thought article c) was computer generated, then you would be right. I can picture you going back a paragraph and read the opening lines again thinking, “Yeah, now that I see it, it makes sense. None of them are Pulitzer Prize material, but c) definitely looks more dull than the others. It must be computer generated”. Somehow your mind has already internalised this fact, and it is starting to reinforce it. If you go back and read them again, I am sure you can spot the flaw right away. As with subliminal messages, once you are aware of them, they do not work any more.

Sorry to disappoint, but you have just been trolled.41 The correct answer is in fact b), that is the computer generated article.42 If you fell for the trick, do not feel too bad. Narrative Science and other companies have many customers in the big media industry that make use of this technology already. Most people just do not notice. The identity of these media firms is secret, but we know they are there because the companies that created these intelligent algorithms have earned several million dollars in a very short time. This software is currently mainly used for sports, finance, business, market, and real estate reporting. I will not go so far as to say that the algorithms developed today can replace all journalists. And I do not expect software to write an editorial about the lack of human rights in China any time soon. But remember, to disrupt an industry you do not need to replace all jobs within it, just a significant fraction.

I have noticed that often people tend to express some form of the following logical fallacy: If you can find one example of a person that cannot be replaced by machines, then the argument of technological unemployment is invalid. On the contrary, I would argue that if you have to rely on that single special example to present your argument in favour of humans, you have just proved my point. The average person within that job type is bound to fall victim to technological unemployment.

Now, just imagine if a few of the big players (e.g. Google, Amazon.com, etc.) that are collecting millions of terabytes of personal information about our reading habits decide to enter the market of automated journalism. We have already seen how Google news has affected news sites by collecting articles into categories and creating personalised news feeds faster and better than any human can. What if these software start to write the stories themselves? How long do you think will that take? If you are thinking decades, you are in for a surprise.

1.6 AI Assistants

You might remember May of 1997, when the legendary chess player Garry Kasparov was defeated by IBM Deep Blue in what has been called “the most spectacular chess event in history.”43 At the time, the plan of IBM was to rely on the computational superiority of their machine using brute force,44 crunching billions of combinations; against the intuition, memory recall, and pattern recognition of the Russian chess grandmaster. Nobody believed it represented an act of intelligence of any sort, since it worked in a very mechanistic way. Boy, we have come so far since then.

The classic “Turing test approach” has been largely abandoned as a realistic research goal, and is now just an intellectual curiosity (the annual Loebner prize for realistic chattiest45 ), but helped spawn the two dominant themes of modern cognition and artificial intelligence: calculating probabilities and producing complex behaviour from the interaction of many small, simple processes. As of today (2012), we believe these represent more closely what the human brain does, and they have been used in a variety of real-world applications: Google’s autonomous cars, search results, recommendation systems, automated language translation, personal assistants, cybernetic computational search engines, and IBM’s newest super brain Watson.

Natural language processing was believed to be a task that only humans could accomplish. A word can have different meanings depending on the context, a phrase could not mean what it says if it is a joke or a pun. One may infer a subtext implicitly or make cultural references specific to a geographical or cultural area. The possibilities are truly endless. A game that captures pretty well the intricacies and the nuances of the English language is Jeopardy! This show that has been on-the-air for half a century and has showcased some remarkable geniuses. Brad Rutter is the biggest all-time winner on the game (making $3,455,102 so far46 ) and Ken Jennings is the record holder for the longest championship streak (74 wins47 ).

In February 2011, IBM’s team decided to take on both champions in a historic match between humans and machine. It was the moment of truth. Watson dominated both humans, bringing home the prize of $1 million (which was donated to charities), while Jennings and Rutter received $300,000 and $200,000, respectively, with both pledging to donate half their winnings to charity. This was a truly historic moment for AI researchers because they were able to reach a frontier that only science fiction writers and futurists believed was possible just a few years ago.

Although IBM’s achievement is impressive, we have to put things in perspective. Watson had access to 200 million pages of structured and unstructured content, consuming four terabytes of disk storage, including the full text of Wikipedia. The hardware is a 2,880 processor cores monster, running on massive parallelism that allows Watson to answer Jeopardy! questions in less than three seconds.48 The total cost of the hardware is about $3 million. Watson’s brain uses 80 kilowatts of electricity and 20 air conditioners,49 while Ken Jennings and Brad Rutter’s brains fit in a shoebox and are powered by a couple glasses of water and a few sandwiches.

Now, I invite you to recall the power of exponential growth in computing. While our brains will remain relatively unchanged for the next 20 years, computer efficiency and computational power will have doubled about twenty times. That is a million-fold increase. So, for the same $3 million you will have a computer a million times more powerful than Watson, or you could have a Watson-equivalent computer for $3.

Watson’s computational power and exceptional skills of advanced Natural Language Processing, Information Retrieval, Knowledge Representation and Reasoning, Machine Learning, and open domain question answering are already being put to better use than showing off at a TV contest. IBM and Nuance Communications, Inc. are partnering for the research project to develop a commercial product during the next 18 to 24 months that will exploit Watson’s capabilities as a clinical decision support system to aid the diagnosis and treatment of patients.50 Recall the example of automated radiologists we mentioned earlier. Watson could be fully capable of performing this task if there was ever the intention of doing so, and even then we would be using only a tiny fraction of its immense power.

This is just the beginning. Watson-like technologies could be used for virtually anything: legal advice, city planning (IBM and Cisco are already working on smart cities),51 and why not policy-making?52

The Internet of Things is coming and we had better be ready. Technology is becoming so cheap and so powerful it will be integrated into everyday objects, which will help us make better decisions. With all objects in the world equipped with minuscule identifying devices, daily life on Earth would undergo a transformation of epic proportions.53 Companies would not run out of stock, nor would they waste products, since involved parties would know which products are required and consumed.54 Mislaid and stolen items would be easily tracked and located, as would the people who use them. Your ability to interact with objects could be altered remotely based on your current status and existing user agreements. We are not quite there yet, but we are getting closer and closer.55

Coming back to the present, let us see what the market has to offer today. Siri is Apple’s attempt to create a personal assistant, and anyone who has ever used it knows that it is little more than a toy. Anybody trying to convince you otherwise is talking marketing trash. Right now it has some built-in AI to recognise speech and create a few connections in the dialogue, make appointments, and send emails; then it queries the computation search engine WolframAlpha to give you computer results to natural language questions; but it does not go very far. The so-called ’smart-assistant’ understands very little of natural language, it does not adapt to many different accents, and it feels nothing like talking to a real person. Generally speaking, it feels like you have to adapt to it, rather than vice versa.

That being said, one cannot overlook its immense potential, given what we learned in the chapter Exponential Technology about the power of the exponential curve. Siri is just the first prototype of a soon-to-be truly smart-assistant that understands any language, spoken by anybody, and helps them with whatever need they might have. In time, it will evolve more and more, becoming increasingly intelligent (meaning useful, not necessarily ’intelligent’ as we are). Its progress will be automatically propagated to all the connected devices, anywhere in the world, instantly. Google is already working on a competitor for Siri as part of its Android platform, and we can expect IBM’s Watson to play a role in the scene as well. And these are just the known players. Today, a team of 3 to 4 people with access to cloud computing can create a revolutionary new intelligent system that can be used by millions of people. The initial investment is very low and the distributed nature of computation allows costs to increase incrementally as the business expands.

We are about to experience tremendous changes in such technologies, the consequences of which are unimaginable for us at the moment. Just as cavemen could not imagine the complex cities and societies we live in today, neither can we anticipate in any accurate detail what is soon to come.

1.7 Autonomous Vehicles

-People often say that something is either obvious and everything will change, or that it will never happen. It turns out things are not quite that simple. Societies are multi-faceted, complex, evolving organisms, with many variables, and a certain degree of unpredictability. Technicians often fail to take into account the human factor, the psychology of the masses, and how events unfold naturally. I think that both perspectives do not really capture the essence of how we, as people, respond to these events. Humanists do not usually understand technology, so their social critique falls short in the face of disruptive change.

Suppose we take the case for automated vehicles. These are self-driving machines: cars, trucks, and buses that do not require a human driver. The idea of self-driving vehicles has been around for a while in popular culture, thanks to science fiction writers. But for the first time, we have the engineering, the mathematical and the computational ability to transform this idea into reality. Some people are enthusiastic about this technology. “It’s about time. I cannot wait to finally get one of those” - said one of the people I interviewed - “It is pretty obvious that human drivers are going to disappear very soon”. But I also received very different answers: “I don’t trust machines, they’ll never be like us. I will never get into a car like that, I want to have control. People won’t accept that, they’ll never have automated cars running on the streets.” This vision is shared by many others I interviewed, some of whom were particularly disturbed by the idea of self-driving cars (surprisingly enough, even young people).

There are many factors to consider, and the evolution of progress goes through various steps. First, there is the development of a new technology. Computer scientists, mathematicians, physicists, and engineers form a small research team somewhere, and decide they want to tackle a specific problem. After a few years of research and development, sometimes even just a few months, they have a working prototype. They test it, improve it, and test it again. They change the conditions, and test it again, and again, until they are satisfied with the result. Now, we have a working technology that has been thoroughly stress-tested under normal as well as extreme conditions, and all the data suggests that this technology is reliable. In fact, it is more reliable than any human; it is safer to use and faster to operate. This represents just the first step. Next comes the social acceptance of such technology. This is not as straightforward as it might seem. Remember that people react very differently to the idea of utilising these machines. Most of the time contrasting opinions are caused by a lack of understanding of the basics of the technology in question. They see it as a matter of trust, or belief. They form their opinion based on intuition, or gut feeling. Whatever the case may be, these different stances are real, and have very serious consequences. As a result, just because a technology exists and it helps us live better, it will not necessarily be adopted right away because of many social factors.

To explain how this process unfolds, I will try to predict what I think is a possible future scenario for the case of self-driving cars. Needless to say, I do not possess the power of precognition, but I will try to make an educated guess. Some of these events, at the time of writing, have already happened. Many have not. Time will tell if I was right or wrong.

1.8 A (possible) History of Self-Driving Cars

Google announced that they have invented self-driving cars. After a few years of research, with very little money, and a small team, they were able to harness the power of machines to solve a very challenging problem of our times. By utilising neural networks and other sophisticated machine-learning algorithms, an immense quantity of data, and thanks to the power of exponentially-increasing technologies that made computation cheaper and faster, as well as sensors, GPS, and laser systems, Google now had a working prototype of a car that drives without the need for a human driver. They then began to test the car on the streets, and let it run for thousands of kilometres. It recognised street signs, traffic lights, pedestrians, dogs crossing, everything around it. It had a 360-degree vision of the surrounding area. It could operate under any conditions, including sun, rain, fog, with icy streets, snowy places, large roads, and small roads. It could move across countrysides, highways, and traffic-intense cities, all while avoiding obstacles. It even prevented accidents from happening when an expected event popped-up, such as a child jumping into the middle of the road, or a bicycle moving into the line without any warning. They then announced these results to the public. People were divided and picked sides quite easily on the spot. Most of them do not bother to investigate: they either love it or hate it a priori. The media did not help either, as many news anchors discard the whole thing with a couple of uninformed remarks, and the public did not receive any information that might change their minds. That is the very reason they watched the news: to become informed. Some news channels provide a very good service; but far too often they instead gave personal opinions, coming from somebody who has no understanding of the subject, and who was paid by the network to display their ignorance and propagate it to the audience.

Meanwhile, further tests were performed and these cars began to gather the attention of many companies and investors. They planned to release the first versions of hybrids, partially-automated vehicles, where the default option is human-driving, but one can switch to automated at any time, and let the car drive for itself. A few states and countries proposed new laws that regulate these cars, insurance companies planned to adjust their policies accordingly. This process took some time, months, and in some cases even years, mostly because social tensions began to emerge. The central issue was safety and responsibility: what if an accident happens, who is responsible? The car owner? The car company? The research team that created the system? The state, which allowed these cars to move freely around their cities? A few brought out another problem: jobs were being taken away by this technology, the displacement of labour (human driver), without a plan to mitigate this loss. These people were largely ignored, and the issue did not come up in the political discourse: if anything, it was the market’s job to fix that problem.

After this media frenzy, the first commercial self-driving cars finally arrived to the market. They could be driven in automated mode only in certain states, so the manual switch option is essential. They are faced with strong opposition by many groups: technophobes, political groups, lobbyists, competitors that did not have this technology yet, or just parents worried for their children, because the news told them that these machines would kill their babies, without any conscience. Acceptance was not easy.

On the other hand, drivers who made use of this technology were extremely satisfied. At the beginning only people with special needs bought the cars (people with reduced mobility and/or vision, the elderly), but then the cars started to gain traction, costs fell, and word of the autonomous car spread all over. Traffic congestion, in states where they allow these cars to drive, are disappearing, and eventually become a thing of the past.56 Owners of the cybernetic cars were very happy about their investment and enjoy the trips. They could relax, read the news, use their smartphone, do some work, or even look outside the windows and enjoy the sky, as if they were on a train. One could simply hop in, choose the destination on the GPS, and enjoy the ride. But the real ’killer-app’ is the “bring me home” command. This is particularly useful in stressful or critical situations. After a long day of work, there is nothing one enjoys more than going home without having to worry about anything. Even more important, they could go out with friends, get drunk, get into the car, mumble “Go home”, or press the big “Home-button” on the dashboard and fall asleep, while the car took care of the rest. Stories of how these cars are helping people and significantly improve the quality of their lives begin to creep in: editorials on newspapers, interviews on TV, and also a few celebrities began to endorse this technology. Traffic congestions decreased, the number of accidents fell significantly. The situation seemed to be changing, and public opinion is now mostly favourable. Then, the first major accident happened.

A self-driving car was roaming around as usual, when another car, driven by a human, crashed into it. The person driving the old-fashioned vehicle was exceeding the speed limit and did not care to follow the street signs either. In short, it was his fault. The cybernetic car tried to avoid the collision, but the other car was simply too fast and it all happened to quickly. The result: the driver of the old car, and his friend next to him, died. News stories went nuts. Headlines like “First self-driving car kills 2 people”, “The killer-machine”, and “Who ’s going to pay for this?” dominate the news arena. The families of the victims are interviewed on national TV, their pain and anger fermented the hatred towards machines that had been dormant up until then. “I knew this would have happened” – “You cannot trust a machine” – “I voted against this law” – “We are going to do whatever necessary to ensure that this does not happen again”, and other nonsense like this was spat out at every corner of the news. Only a few brought out the fact that, between the time self-driving cars and the first major accident happened, thousands of accidents among human drivers occurred, where hundreds died and none of them made it into the news. It did not matter: facts are not important, what matters is our perception of reality. Some states declared that they will never allow these infernal machines to do any more damage, and refused to accept them. More legislation, more public discourse, more debates and opposition soon followed.

Meanwhile, technology advanced exponentially: cars became even more reliable, they required less energy, their algorithms improved. They were cheaper, widespread, more companies developed such technologies, and demand for these cars rose. Soon, it became the only growing market in the automobile industry, and companies that failed to innovate risk dying off. On the other hand, there is a small group of dedicated individuals who spoke about the pleasure of driving, the value in keeping your mind occupied and the “good old days”. Also, they said it was important to have control over our tools, and that the direction people were taking was ugly and dangerous. The had a few supporters, and they remained faithful to this view, regardless of the ever-growing advancements on the field.

After a few years, these cars were widespread across most developed countries, they were still hybrid models, but people relied on their driving skills less and less. Streets became more secure, and traffic jams were greatly reduced. Some bold companies began to design entirely new car concepts: fully autonomous, cybernetic vehicles, where the human driver is no longer needed. As such, they could redesign the cabin from the ground up. Seats could now move in any direction, all four people could face each other if they liked, in circle. Being in a car now became a whole different experience; it could be a truly social event. Given the situation, one would expect every car, bus, truck, and taxi to run autonomously by now. It would certainly have been the right choice: more efficient, less accidents, no traffic jams, cheaper and more reliable than human drivers…having autonomous vehicles would be logical. But things do not always go according to what is logical. They follow complex dynamics that have to do with society, group thinking and complex dynamics that have little to do with technology and what is good; and a lot to do with politics, marketing, emotional attachments, old habit, delusions, beliefs, and what appears to be good.

The invention and creation of a technology may be a hard problem, but sometimes social acceptance of that technology is a much harder one.

Notes

1 According to the Japan Vending Machine Manufactures Association website, there are 8,610,521 vending machines in Japan, or one machine for every 14 people.
http://www.jvma.or.jp/information/qa_01.html

2 Amazon buys army of robots, Julianne Pepitone, 2012. CNN Money.
http://money.cnn.com/2012/03/20/technology/amazon-kiva-robots/index.htm?hpt=hp_t3

3 Tesco Homeplus Virtual Subway Store in South Korea.
http://www.youtube.com/watch?v=fGaVFRzTTP4

4 The Weight of Walmart (Infographic)
http://frugaldad.com/2011/12/01/weight-of-walmart-infographic/

5 Strikes End at Two Chinese Automotive Suppliers, 2010. Reuters.
http://www.reuters.com/article/idUSTRE66L0A220100722

6 Table 3. The Circuits Assembly Top 50 EMS Companies, 2009. Circuits Assembly.
http://circuitsassembly.com/cms/http://robotswillstealyourjob.com/sites/robotswillstealyourjob.com/files/book/stories/Articlehttp://robotswillstealyourjob.com/sites/robotswillstealyourjob.com/files/book/1003/1003buetow_table3.pdf

7 Forbes Global 2000: The World’s Biggest Companies – Hon Hai Precision Industry, 2010. Forbes.
http://www.forbes.com/companies/hon-hai-precision/

8 Which is the world’s biggest employer?, 2012. BBC News.
http://www.bbc.co.uk/news/magazine-17429786

9 Apple partnership boosting Foxconn market share, 2010. CNET.
http://news.cnet.com/8301-13579_3-20011800-37.html

10 Foxconn to replace workers with 1 million robots in 3 years, July 2011. Xinhuanet News.
http://news.xinhuanet.com/english2010/china/2011-07/30/c_131018764.htm

11 Companies Making The Necessary Transition From Industrial To Service Robots, 2012. Singularity Hub.
http://singularityhub.com/2012/06/06/companies-making-the-necessary-transition-from-industrial-to-service-robots/

12 emphFoxconn Factories Are Labour Camps: Report. South China Morning Post.

13 Foxconn Security Guards Caught Beating Factory Workers, 2010. Shanghaiist.
http://shanghaiist.com/2010/05/20/foxconn-security-guards-beating.php

14 Revealed: Inside the Chinese Suicide Sweatshop Where Workers Toil in 34-Hour Shifts To Make Your iPod, 2010. Daily Mail (London).
http://www.dailymail.co.uk/news/article-1285980/Revealed-Inside-Chinese-suicide-sweatshop-workers-toil-34-hour-shifts-make-iPod.html

15 Suicides at Foxconn, 2010. The Economist.
http://www.economist.com/node/16231588

16 Canon Camera Factory To Go Fully Automated, Phase Out Human Workers, June 2012. Singularity Hub.
http://singularityhub.com/2012/06/06/canon-camera-factory-to-go-fully-automated-phase-out-human-workers/

17 China Is Replacing Its Workers With Robots, 2012. Business Insider.
http://www.businessinsider.com/credit-suisse-chinese-automation-boom-2012-8

18 The Machines Are Taking Over, Sep. 14, 2012. The New York Times
http://www.nytimes.com/2012/09/16/magazine/how-computerized-tutors-are-learning-to-teach-humans.html

19 Why Software Is Eating The World, 2011. The Wall Street Journal.
http://on.wsj.com/pC7IrX

20 In the TV series Star Trek, a replicator works by rearranging subatomic particles, which are abundant everywhere in the universe, to form molecules and arrange those molecules to form the object. For example, to create a pork chop, the replicator would first form atoms of carbon, hydrogen, nitrogen, etc., then arrange them into amino acids, proteins, and cells, and assemble the particles into the form of a pork chop.
http://en.wikipedia.org/wiki/Replicator_(Star_Trek)

21 Will 3D Printing Change The World?, 2012. Forbes.
http://www.forbes.com/sites/gcaptain/2012/03/06/will-3d-printing-change-the-world/print/

22 Objet Connex 3D printers.
http://www.ops-uk.com/3d-printers/objet-connex

23 iPhone 4’s Retina Display Explained, Chris Brandrick, 2010. PC World.
http://www.pcworld.com/article/198201/iphone_4s_retina_display_explained.html

24 3D printing.
http://www.explainingthefuture.com/3dprinting.html

25 A primer on 3D printing, Lisa Harouni, 2001. TEDSalon London Spring 2011.
http://www.ted.com/talks/lisa_harouni_a_primer_on_3d_printing.html

26 3D-printed prosthetics offer amputees new lease on life, 2012. Reuters.
http://www.reuters.com/video/2012/02/27/3d-printed-prosthetics-offer-amputees-ne?videoId=230878689

27 3D printer used to make bone-like material, 2011. Washington State University.
http://wsutoday.wsu.edu/pages/publications.asp?Action=Detail&PublicationID=29002&TypeID=1

28 Making a bit of me, a machine that prints organs is coming to market, 2010. The Economist.
http://www.economist.com/node/15543683

29 Transplant jaw made by 3D printer claimed as first, 2012. BBC News.
http://www.bbc.com/news/technology-16907104

30 What drives us. Bespoke.
http://www.bespokeinnovations.com/content/what-drives-us

31 Thingiverse.
http://www.thingiverse.com

32 First Downloaded and 3D Printed Pirate Bay Ship Arrives, 2012. TorrentFreak.
http://torrentfreak.com/first-downloaded-and-3d-printed-pirate-bay-ship-arrives-120205/

33 30-storey building built in 15 days Construction time lapse. YouTube.
http://www.youtube.com/watch?&v=Hdpf-MQM9vY

34 Time lapse captures 30-story hotel construction that took just 15 days to build, 2012. The Blaze.
http://www.theblaze.com/stories/time-lapse-captures-30-story-hotel-construction-that-took-just-15-days-to-build/

35 Annenberg Foundation Puts Robotic Disaster Rebuilding Technology on Fast Track, 2005. University of Southern California School of Engineering.
http://viterbi.usc.edu/news/news/2005/news_20051110.htm

36 House-Bot, December 30, 2005. The Science Channel.

37 Census of Fatal Occupational Injuries Summary, 2010. Bureau of Labour Statistics.
http://bls.gov/news.release/cfoi.nr0.htm

38 Caterpillar Inc. Funds Viterbi ‘Print-a-House’ Construction Technology, 2008. University of Southern California School of Engineering.
http://viterbi.usc.edu/news/news/2008/caterpillar-inc-funds.htm

39 Colloquium with Behrokh Khoshnevis, 2009. Massachusetts Institute of Technology.
http://www.media.mit.edu/node/2277

40 GSP-09 Team Project: ACASA, 2009. YouTube.
http://www.youtube.com/watch?v=172Wne1t_2Q

41 Problem?
http://www.urbandictionary.com/define.php?term=trolling

42 Are Sportswriters Really Necessary? Narrative Science’s software takes sports stats and spits out articles, Justin Bachman, 2010. Newsweek.
http://www.businessweek.com/magazine/content/10_19/b4177037188386.htm

43 Garry Kasparov vs. Deep Blue, Frederic Friedel. Daily Chess Columns.
http://www.chessbase.com/columns/column.asp?pid=146

44 In computer science, brute-force search or exhaustive search, also known as generate and test, is a trivial but very general problem-solving technique that consists of systematically enumerating all possible candidates for the solution and checking whether each candidate satisfies the problem’s statement. For example, a brute-force algorithm to find the divisors of a natural number n is to enumerate all integers from 1 to the square-root of n, and check whether each of them divides n without remainder.
http://en.wikipedia.org/wiki/Brute-force_search

45 Chatbots fail to convince judges that they’re human, 2011. New Scientist.
http://www.newscientist.com/blogs/onepercent/2011/10/turing-test-chatbots-kneel-bef.html

46 Did you Know?, Jeopardy!
http://www.jeopardy.com/showguide/abouttheshow/showhistory/

47 Computer Program to Take On ’Jeopardy!’, John Markoff, 2009. The New York Times.
http://www.nytimes.com/2009/04/27/technology/27jeopardy.html

48 According to IBM, Watson is a workload optimised system designed for complex analytics, made possible by integrating massively parallel POWER7 processors and the IBM DeepQA software to answer Jeopardy! questions in under three seconds. Watson is made up of a cluster of ninety IBM Power 750 servers (plus additional I/O, network and cluster controller nodes in 10 racks) with a total of 2880 POWER7 processor cores and 16 Terabytes of RAM. Each Power 750 server uses a 3.5 GHz POWER7 eight-core processor, with four threads per core. The POWER7 processor’s massively parallel processing capability is an ideal match for Watson’s IBM DeepQA software which is embarrassingly parallel (that is a workload that is easily split up into multiple parallel tasks).
http://www-03.ibm.com/systems/power/advantages/watson/index.html

49 Instant Reaction: Man-Made Minds, David Ferrucci, 2011. World SCience Festival.
http://worldsciencefestival.com/blog/instant_reaction_man_made_minds

50 IBM’s Watson heads to medical school, Nick Wakeman, 2011. Washington Technology.
http://washingtontechnology.com/articles/2011/02/17/ibm-watson-next-steps.aspx
Wikipedia, Watson.
https://en.wikipedia.org/wiki/Watson_\%28computer

51 Mission Control, Built for Cities. I.B.M. Takes ‘Smarter Cities’ Concept to Rio de Janeiro , Natasha Singer, 2012. New York Times.
http://www.nytimes.com/2012/03/04/business/ibm-takes-smarter-cities-concept-to-rio-de-janeiro.html?pagewanted=all

52 Will IBM Watson Be Your Next Mayor?, 2012. Slashdot.
http://yro.slashdot.org/story/12/04/27/0029256/will-ibm-watson-be-your-next-mayor

53 Computers to Acquire Control of the Physical World, P. Magrassi, A. Panarella, N. Deighton, G. Johnson, 2001. Gartner research report. T-14-0301.

54 A World of Smart Objects, P. Magrassi, T. Berg, 2002. Gartner research report. R-17-2243.
http://www.gartner.com/DisplayDocument?id=366151

55 The Internet of Things. Wikipedia.
http://en.wikipedia.org/wiki/Internet_of_Things

56 Study: Intelligent Cars Could Boost Highway Capacity by 273%, 2012. Institute of Electrical and Electronics Engineers.
http://spectrum.ieee.org/automaton/robotics/artificial-intelligence/intelligent-cars-could-boost-highway-capacity-by-273


You're reading the free version of a book that required more than one year of work. You can show your support by purchasing a copy on Amazon (Kindle/Paperback), iBookstore, PDF or ePub direct download. All files are rigorously DRM-free. Want to support my work, but don't want to buy the book? Send a donation via Bitcoin (1LeUniaxmEqYh6WJQ9ktuSYonkLG5HvcbZ) or PayPal :)

Get new chapters every week for free!