The Future of Artificial Intelligence And Humanity In A World Of Unemployment: An Economic and Philosophical Analysis
Table of contents:
Introduction
• What is intelligence?
Narrow AI
• Automation
• Self-driving cars
o The Economics of Self-driving cars
o The Ethics of Self-driving cars: Two thought experiments
• Professional jobs, Creative jobs, The future of education
• Luddite fallacy
Strong AI
• How should we treat human-like intelligence?
Super AI
• The Singularity
o How a robot behaves
o What we can do
• Do we really want such a future?
o What’s so good about immortality?
o Lonely planet: Colonising space
o When minds become computers: What is the self?
Emotional AI
Conclusion
Chapter 1: Artificial Narrow Intelligence
When most people are confronted with the question of what is AI, the answer may well often be a humanoid machine whose function closely mirror humans – a concept that has been effectively advocated by Hollywood. This is because it is mostly through fictional platforms that the populace is exposed to and interacts with the idea.
In reality though, artificial intelligence is, at least in the present moment, much more abstract and commonplace than most people realise. Indeed, most people, especially those who work in the modern financial sector, have seen it at work. More specifically, they would be seeing what robot experts call Artificial Narrow Intelligence, or sometimes Weak AI.
It is the software that scans your train ticket and makes sure you are on the right platform, the algorithms used to arrange your Facebook timeline and filter your email, the impressive programming that helped a machine beat humans in a game of chess. On a larger scale, AI is used to regulate traffic lights, control the direction of the lift and even electrical power grids. These are all examples of Weak AI. Society, economies and stock markets have grown very much dependent on this system.
None perhaps recognise this truth better than multinational corporations and stockbrokers. Firms now largely compete in terms of who can launch the most groundbreaking technology into the market first. Trading floors used to be floundered with human brokers buying and selling assets every second, but now it is not so much a human endeavour, but a machine-operated one. As business and institutions are spurred on to make the most cost-effective decisions, the trend will see increasing integration of Narrow AI with daily life. The first area to be affected is probably labour.
Automation v Traditional labour:
Automations are robots. They have great mechanical flexibility, but more or less simple Artificial Intelligence. In the present time, robots in factories are nothing new. When putting together a car, automation can outstrip any human worker in terms of power, pace and precision. After all, these mechanical arms do not grow sleepy or tired, and instead of demanding minimum wage they cost only the price of electricity and the occasional maintenance fee.
However, they have not taken over all forms of human labour because they are only cost effective in narrow situations (hence the name Narrow AI). This older version of automation is remarkably specialised. But recent years have seen advancement of new kinds of automation that are general-purpose, which means they can learn to do any task we humans want them to learn by watching us do it. An excellent example of this is Baxter.
Baxter has two mechanical arms, which he uses to perform tasks. He has a screen with two eyes on it to indicate the direction he was looking at. If you grab Baxter’s arm, he brings his head around, signaling that he is ready to be trained. Unlike previous specialised versions, Baxter is not programmed for one specific job. In demonstration videos, his creators from Boston-based company Rethink Robotics showed how Baxter can be taught to fold a simple T-shirt, brew coffee or pour a glass of champagne. Baxter’s sensors also allow him to know when there are people nearby, giving him the ability to adapt to the environment. He is also able to express confusion when something is not right, and even surprise and sadness.
There is little doubt even today that the technology behind Baxter is remarkable indeed. But Baxter is slow to work, in the sense that he cannot multitask as quickly as humans. Is it still efficient to have a machine like Baxter around?
The answer is probably yes.
Take this example of the early computer. At one time, these machines were hugely expensive, taking up entire basements or warehouses. The old models had once been painstakingly slow. But when circumstances allowed them to be produced more cheaply, computers quickly became vital to almost every aspect of human activity. This growing appetite (demand) for computers is what drives them to be both more powerful and affordable each year. Likewise, a multi-purpose robot like Baxter today may still have its limitations; but Moore’s Law and the exponential technological growth means that robots like him can improve on a digital timescale. In other words, fast. Once production cost has sufficiently dropped, many more people will have heard of Baxter. And then they just might think ‘Well, it couldn’t hurt to have a bot like that…’
The first places to receive Baxter-like robots will be small-chain manufacturing businesses. His adaptability is great for his employers, who must no longer spend huge sums to supervise his production.
His blue-collared colleagues, though, may not be so happy. Their job now hangs in the balance. They may find that their labour has become dispensable and they themselves possibly unemployed through no fault of their own. It will not happen within a fortnight, but in the medium to long term, demand for low-skill labour will likely shrink.
It is tempting to believe that the widespread impact of narrow AI will only stop at industrial proletariat, but such belief would be greatly misinformed. In recent years, self-checkout machines have taken over jobs at the till, which in the past would number in the thousands. You would not be able to find a single large supermarket chain in the United Kingdom today that does not make use of these automatic cashiers. The same goes for the hundreds of thousands of baristas employed worldwide. It is now possible to find barista robots in small kiosks that offer a range of lattes, cappuccinos, espresso and café au lait. The robot will take your order and memorise the exact way you would like your coffee made. Gone are the days one had to drive to one single coffee stand to get that specially made, gourmet hot chocolate. Using this new system, you can obtain that luxurious drink no matter where you are, in any kiosk.
What about jobs in hotel room service, butlery, or washing and ironing clothes in hospitals? IEEE Spectrum demonstrates how worldwide research and development there have already created robot prototypes just for that.
The point here is that invention and increasing availability of these technologies spells out possible unemployment for hundreds of thousands of low-skill workers.
You might ask if computing technology has already reached such a level, why are we not seeing robot cleaners vacuuming our homes? An explanation for that is that these machines are still in its early, experimental stages. They are the results of years of research and computing brilliance, so it is not surprising that they are still too expensive for mass production.
However, a number of things we take for granted today was also not affordable for most people in the past. Yet, seeing how widespread personal computers had become over the last decade once production techniques has caught up, we can already imagine how in the future, rare technologies today may turn commonplace.
Self-driving cars: Economic Trade-off and Ethical Dilemmas
The Economics of Self-driving cars
In Jan 2016, the BBC first published an article about the plans to introduce driverless cars onto London streets in the near future, based on the same design as the electric shuttles (also self-driving) used in Heathrow Airport.
Westfield Sportscar, the company responsible for testing and manufacturing these driverless pods, will not be the first to experiment on this seemingly futuristic idea. Google’s ambitious self-driving car project already had its vehicles traverse hundreds of thousands of miles on Californian roads and coasts, which may give valuable insight into the future of data-driven information.
It would not be surprising if a part of the public would skeptical of the idea of their vehicles driving them, of not being in control behind the wheel. But while these reservations are understandable, statistics allow a glimpse of the heavy toll resulting often from simple, human mistakes and paid in lives. A study into the causes of traffic accidents published in 1979 found that "human errors and deficiencies" were a definite or probable cause in 90-93% of the incidents examined. Put in simpler terms, it is chiefly human error that was the primary cause of traffic accidents.
The Department of Transport estimated that 1,713 people were killed in reported traffic accidents every year in the UK. While numbers have leveled off in recent years, road accidents still cost 1.25 millions lives worldwide each year and costs government roughly 3% of GDP. All in all, traffic accidents incur billions of dollars per annum in healthcare costs, lawsuits and compensation.
This loss of human lives can be prevented if self-driving software are allowed to take over, or so the argument goes. Unlike human drivers, machine intelligence do not get distracted by phone calls, are not influenced by mood or emotions, never suffer from insomnia and certainly do not drink -drive. They are precise in a way humans often are not. These cars have 360-degree visibility and 100% attention out in all directions at all times. Its sensors can detect other vehicles, cyclist and pedestrians at a distance.9
But auto-vehicles are not infallible machines. In fact, over the six years since Google initiated the Self-driving Car project, their vehicles had been involved in eleven minor accidents. However, there were no injuries and parties only sustained light damage. In none of these incidents were the self-driving cars the cause of the accident. Seven out of the eleven minor incidents involved other drivers hitting the car from behind. Again, it had been human error that had caused the accidents.
The goal is not to create a vehicle completely free from error – it would be unrealistic. However, self-driving cars only have to make fewer mistakes than humans do for it to be more efficient than manual driving. Trials already indicated that to be the case.
Besides whom this most directly concerns (the driver), it would also be within policy-makers’ interest to promote self-driving cars. The average driver in the US alone spends a mean of 52 minutes per day in commute. That is four billion hours10 for the whole nation every year, spent passively driving when it could be used for much more productive work. It also equates to 2.4 billions of gallons of gasoline squandered while waiting in traffic.10 All in all, public and private expenditure associated traffic problems are both economically inefficient and a negative externality – a thing society is better off without.
And much like Baxter, a future where the robotic car dominates the streets will spell out the end for many jobs in the transportation industry, in fact any industry that relies heavily on shipping and transportation. The AI behind this car has implications that extend beyond simply getting us humans across the map – it has the potential to replace much of the current transportation system. Smaller versions of these machines can move goods around warehouses while larger ones can carry ores and mineral deposits in mines. In the future AI might be capable of navigating across oceans, at which point shipment will start to be automated.
This technology alone targets 19 million people are employed in the mining, construction and manufacturing industry in just the US , or an extrapolated 63 million jobs globally in only those industries. This spells potential mass structural unemployment that will possibly shake the very foundations of the global economy.
But there is a saying that “one man’s loss is another man’s gain”. Employers and managers may find nothing to be unhappy about in this new situation. After all, costs have fallen substantially and it is much easier to balance the books and write reports.
Some argue that shareholders and company directors may not necessarily want to lower price even as costs fall. Firms, after all, are mostly self-interested agents. Any company would have an incentive to protect its own economic profit. However, it is important to bear in mind that not all firms want to maximize profits in the short run. In contestable market conditions, any firm that can drive down its costs and exploit economies of scale is likely to have a price advantage over its competitors so that eventually, real price for goods and services may drop.
The role of transportation in economic theories is highly evident. It is essential to company pricing strategies, economic mobility and free trade. As the world sees increasing demand for quick and efficient transport systems, automation can help fill in the shortage gap and help consumers worldwide to enjoy a much richer variety of desirable goods and services. Falling postage fees also lessons economic inequality between industrialised and developing countries, as well as on regional and local scale.
In the next section we explore the ethical issues surrounding the Self-driving car. The intention here is not to undermine the impact of potential unemployment for workers in these industries and their families. Losing a job and a source of income and stability is debilitating for anyone, but for the sake of the argument, we must move on.
The Ethics of a Self-driving car: Two thought experiments
As impressive as autonomous vehicles may be, it does raise some concerning ethical questions. Seeing that self-driving cars operate on algorithms and computer programmes, it means it is not unsusceptible to software failure. If it misinterprets a worn road sign, is the government department of transportation answerable? And what of cyber security? Can the software of the car be hacked? Can others track down your movements if there is some breach in data protection? But most importantly, who is liable for an accident when it happens?
Imagine one day, you are cruising down the streets in your auto-car and are hemmed in on all sides. There is a motorcyclist on your right and another car on your left, when heavy containers start to fall off the lorry in front of you. Everything was so sudden the auto-car could not stop in time to avoid an accident. What does it decide?
If it keeps moving forward, it will hit the falling object and you, the driver will be killed. By doing this, it minimises dangers to others, but compromises the driver’s safety. If it swerves right, the car prioritises your safety, at the cost of the innocent motorcyclist’s life. If it swerves to the left and hit the other car, which has a high passenger safety rating, then it has neither minimised harm to others nor has it prioritised your own safety (though at least both drivers have a higher chance of surviving the accident).
It is true that these scenarios are unlikely happen in real life. Yet, much like Foot and Thomson’s trolley problem, it serves well as a thought experiment – designed to test our intuitive response on ethics and provoke our thoughts.
Some think that, in such a technologically advanced future, the problem in the mentioned scenario could be averted by making the vehicles communicate; so that the left car can drive out of the way and avoid a collision with your car when it swerved. However, this solution is only reliable if all cars on the road were automated to an extent. But it is never guaranteed that everyone will be strapped to an autonomous car. Smartphones today are commonplace, but even so they are not affordable for some people. And what if in another scenario, instead of a car, now you are surrounded both sides by motorcyclists, only one of whom is wearing a helmet. What will the car decide this time?
At first, it may be logical to choose the lesser evil – hit the motorcyclist with the helmet to minimise harm. But by doing so, are you not effectively punishing this person for conscientiously wearing a helmet? But by deciding to hit the other motorcyclist, the auto-car would have overstepped its programming and violated its design to minimise harm!12
It is clear that auto-cars shift the responsibility from driver to designer. Had the car been manually driven in both these two experiments, any move to avoid the collision would have been a subconscious reaction. But an auto-car functions based on premeditated algorithms, and it will discriminate a type of object to crash into. If the programmers were to instruct the car to do the same, they would be making a deliberate decision of whom they would rather be sacrificed if necessary.12 Even if the driver was deliberately veering the car, it is still one thing to make a choice yourself, and another for a machine to make the decision for you.
Professional and Creative Jobs
It is easy to look at technology now and t
The Effects Of Automation On Current Economic Models
If automation starts taking over David Ricardo’s labour theory of value states those differences in comparative cost/advantage are primarily derived from differences in the productivity of labour. This theory has continued to hold true for the most part until now, but it might be rendered obsolete in a time when discrepancies in labour productivity no longer mattered so much as machine productivity.
Of course, there are other factors that decide a country’s opportunity costs, but say the Ricardian labour theory
How robots behave
As synthetic intellects entered the domains previously exclusive to humans, they are likely behave in a way that seems to disregard morality or awareness of side effects, and which people in society find increasingly distasteful. The reason for their lack of a sense of ‘fair play’ is also the reason computer minds are so efficient – they are designed to achieve singular goals. Digital advertising companies like Rocket Fuel devise elaborate AI to influence consumer behaviour. Investors at the New York Stock Exchange use High-Frequency-Trading programmes to process millions of transactions and get the upper hand of the financial systems, by working out the bare minimum of what they had to pay for what they want and charge the most they can to maximize profits.
But by that extension, synthetic intellects may decide that the most efficient way to prepare things in case of an emergency is to snatch all the groceries before a big storm, even if it meant leaving none for others. You may instruct your self-driving car to find a parking space for itself, and it does that by stealing the place someone else had been waiting for. Programmers may provide the algorithm to work out a way to make humans smile, and the AI might decide that the most effective way to achieve this goal was to control facial muscles in humans via sticking electrodes and cause them to contort into smile.