The goal of this column is to share knowledge and thoughts about AI (Artificial Intelligence) as it is an evermore-important factor in day-to-day life. Be it in your free time or at work, AI is there. Netflix uses it to determine what movies match your taste. Companies use it to analyze trends and solve their problems. AI is everywhere but – as you might say – it is very complicated to understand it and oftentimes, you don’t even have to (you don’t have to know how Netflix works to use it). Still, the more you know about it, the more control you have over it and the less you will feel overwhelmed by it. This article isn’t about in-depth knowledge about how to code AI or how to use it – we’ll leave that to the specialists and the enthusiasts – but more about the general concepts you need to know about regarding AI and about how AI is affecting your live. This knowledge might help you understand the implications of an increased use of AI and an increase in AI-capabilities. This will be useful both to understand the business world and to have a better grasp of some of the big trends that shape society.
Before we get to that, let’s start with the basics: what is AI? To begin answering this question, it is worth taking a look at the cycle of conferences set up by emlyon business school‘s AIM institute – AIM stands for Artificial Intelligence in Management. It is a “multifaceted initiative focused on understanding the opportunities and implications of artificial intelligence for the management of organizations, industries, and business ecosystems”.
AIM conducts research about the impact of AI for work and value creation, with the goal of bringing about innovation in the field. It also tries to create some pedagogical material for students and the general public in order to educate them about the topic at hand. To start giving a definition of AI we will compare two types of AI using AIM’s conferences as a support.
The different types of AI
One of AIM’s conferences at emlyon business school was about a central topic of the development of AI: machine learning (you can find it rebroadcasted on YouTube). We won’t get into that right now but instead will focus on the different types of AI that were mentioned and on some examples of AI systems. The speaker, Celine Robardet (professor working at INSA Lyon –CNRS) started by going back to the beginnings of AI and using the example of computers playing chess. First of all, it is important to know that AI mimics human processes.
AI tries to artificially reproduce a cognitive process that is all too human – that is, it tries to acquire knowledge and understanding through thought-experiences and the senses. It then uses that information to perform a task. At first, AI was task-specific: it could assist with or take over one specific task but without any carryover to other tasks. For example, a computer taught to play chess can become good (even very good) at chess, but won’t be able to use that knowledge and understanding in other fields. It wouldn’t be able to play checkers for instance. The Computer would have to learn everything from scratch without being able to use its chess-playing skills to reduce the learning curve of subsequent games. These systems are called narrow AI.
Narrow AI is narrow in the sense that it is only able to work in a narrow, pre-established field. During the conference several examples of narrow AI were given, as regards chess playing.
The first one was Deep blue, the computer that eventually beat Gary Kasparov – the chess world champion at the time. It took Deep blue two tries as in 1996, the best human chess player at the time was still able to “beat the machine” in a 6-match tournament. It finally succeeded in its task in 1997, when Deep blue was able to beat Kasparov wining 3,5 out of the 6 matches they played. It is a big win for AI. It is safe to say that at the time, AI could already match top-level chess players. It used a lot of computational power to do so. Instead of trying to learn how humans played chess, Deep blue looked at every possible move and through calculation, tried to determine the best possible move. This required a lot of processing power, so much so that this technique couldn’t be used for chess before the 1970’s. The version of Deep blue that beat Kasparov was 1,80m tall and weighed 1,4 tons.
Now, with the progress of technology, the miniaturization of components and the drastic increase in processing power, brute force computation can go much further. This is one of the reasons why computers now consistently beat the best human chess players. So much so, that the best chess player is now a machine. Narrow AI is not the only form of artificial intelligence though. Other AI systems go further and are sometimes considered as even more “intelligent”.
General AI is capable of transferring its knowledge and understanding from one domain to another. This type of AI isn’t necessarily more powerful – meaning that it doesn’t have to have more processing power. One could have an incredibly powerful narrow AI, set on a single task and a less powerful general AI being able to work on different tasks. It is a different kind of AI. As some might argue, it is the “scary kind”, because it is closer to a dystopian scenario – such as the one portrayed in the Terminator movies or in Isaac Asimov’s I Robot – where a machine becomes super intelligent, takes over all tasks and then develops a mind of its own. Let’s see how much truth there is to that and how realistic those scenarios are.
As regards a machine transferring its knowledge from one field to another, this is possible but we are far from what happens in those movies. The power of general AI is still limited: AI researchers aren’t there yet – at all. One cause of concern might be that the progress of AI is exponential. Computers become more and more powerful as explained by Moore’s law: every year the number of components per integrated circuit doubles, which more or less means that every year processing power doubles.
As a result AI-research can go further. AI is also becoming more and more proficient at performing tasks (through machine learning and deep learning as we will see in the next article). This might be especially worrisome in the cases where computers/AI teaches itself. Facebook made the news in 2017 when two of its AI-bots, set-up to communicate with each other, started using a language that the researchers couldn’t understand. They also started to “negotiate” with one-another and to make “deals” for the future. Facebook researchers had to shut down the project. This shows that science fiction might still be distant but that one has to be careful now before it is too late.
That being said, examples of general AI are still very limited. Even tools such as IBM Watson are viewed as narrow AI and not general AI or at least not a strong version of general AI. That’s saying a lot since these AIs can already do a lot. IBM Watson can be used in many fields as a search engine, to answer specific questions. It was even by a company named ROSS used as a base to build a so-called AI-attorney – a tool to help (and eventually replace?) legal research assistants, digging through court documents, law books, case files and so on.
AI: cause for concern or celebration?
The risks of AI
The thought of General AI is cause for concern for many people. Some, such as Elon Musk, advise caution regarding the unsupervised development of AI. The idea is that, as we have seen with Facebook’s experiment, AI can escape human understanding and thus human control relatively fast. This might be worrisome as it isn’t self-evident that progress of AI beyond human-capabilities is beneficial for society. It may be the case, that AI goes too far, as one can see in the many science fiction books or movies.
Others look at the potential problems caused by AI differently. In his book 21 lessons for the 21st century, the historian Yuval Noah Harari looks at the future of AI and its impact of society mainly through the job-market. First, although he does not exclude this possibility, he advises not to look at artificial intelligence as something being able to develop consciousness: to feel and want things. A machine becoming “smarter” in the sense that it can solve evermore complex problems doesn’t necessarily mean that it will develop a mind of its one, with its own set of goals.
This line of reasoning would thus exclude scenarios such as the one in the HBO series Westworld, were AI gradually inches forward on the road to consciousness. Real-life cases such as Facebook’s bots might lead to different conclusions – although it is hard to determine what really happened.
Harari focuses much more on the consequences of the progress of AI without consciousness – be it narrow or general AI – and its implications on the job-market. As AI can perform more and more tasks better than humans, jobs will inevitably change or disappear. According to Harari there is a risk that exponential progress in AI technology – even if it is done under human supervision and under human control – will lead to mass unemployment. It may thus lead society to what he called a “useless class”: people with nothing to offer of the job market. This would affect not only blue-collar but also many white-collar jobs.
Others share his concern but it is important to get the details right. In a Time’s article, Kai-Fu Lee author of AI superpowers: China, Silicon Valley and the new world order, looks at which jobs will disappear in the future. According to him AI, through a method known as deep learning, will outperform humans on jobs with lots of available data and where the task at hand is repetitive/routine. That might be most of the available jobs – especially if self-driving cars take over truck and taxi driving. According to him, 4 types of jobs are not threatened by AI: creative jobs, complex strategic jobs, “as-yet known jobs” (the jobs created by AI) and empathic or compassionate jobs (such as teaching or nursing). It is hard to say how many jobs AI will create. Many fear it won’t be enough to offset the number of jobs that might disappear. As for creative and complex strategic jobs, they aren’t enough of them, they aren’t for everyone be it because of skill-set or preference.
One silver lining might be that compassionate jobs are less threatened and that they might even be valued more and more. In a world were AI and machines do all the heavy lifting, some human contact and compassion might be much wanted.
The benefits of AI
It is often easy to forget the other side of the coin. Though AI is undoubtedly scary in some aspects and might have unintended negative consequences, it also has a lot to offer. That being said, it is often hard to know specifically what AI has to offer, partly because it is so complex and partly because we sometimes take it for granted. For a phone to become a smartphone, some AI has to be involved. Most people wouldn’t want to trade-in their smartphone for a “dumber version”. What about AI in general?
The world underwent many technological transformations in the past. Looking back, it is easy to see how innovations such as the printing press, the combustion engine or the agricultural revolution made the world more prosper. Looking at AI today it is difficult to be as enthusiastic. This may be for two reasons. Firstly, AI may be a different kind of technological revolution, one that could lead us to more problems down the road instead of bringing more prosperity to the world. Some of those concerns were underlined just before, many of which are credible. Secondly, it is harder to see the benefits of something that is still knew and thus not fully known to us and understood by us. We might be judging AI more harshly than it deserves because we want to stay on the side of caution. Although that isn’t necessarily a bad strategy, going forward we also need to objectively look at the benefits of AI.
One conception is that AI will enable us to find better solutions to many societal problems. Given enough data, AI is already far more capable than humans to perform complex data analysis and computation-heavy tasks. It is a powerful tool. Used in the right hands with the right intentions it might help us find solutions to problems that remained unsolvable before. AI can for instance help companies optimize and automate their operations. Though this often means job-losses for workers, it also means higher productivity, cheaper and better products. Regarding the job market, AI will create job as well. Qualified people will be needed to harness the benefits of this new technology. In many cases this may lead to fewer jobs overall.
All job-losses can’t be laid at the feet of AI and automation though. In western countries much of the job losses in the industrial sector come from international competition with cheap labor and thus lower costs.
If factories close in Europe or in the US it has more so to do with China than with the progress of technology. On the contrary, a part of western countries’ industry stayed competitive through automation and technologies such as AI. It is also worth noting that international competition, although it led to job losses in many Western countries, led to increased growth an value creation all around the world. Globalism undoubtedly has its flaw and needs to be regulated, but it made the world richer and lifted millions of people out of poverty. Depending your frame of reference you can arrive at different conclusions. Looking at workers who lost their jobs because of competition from China, it’s hard to say Globalism made things better. Looking at the big picture it did make things better, especially looking at the alternatives.
The same thing can be said for AI. Depending on whom you ask and were you look, AI is either an incredible opportunity or a doomsday scenario. In many job sectors automation and tools such as AI enable workers to do their job better. It removed a lot of risk factors and some exhaustive tasks. On the mater of saving jobs and curbing unemployment many see AI as a threat. Others consider that it is more important to take care of workers than jobs. AI might destroy some jobs but create others – the question is how many. AI is a powerful tool to make the world richer. Few doubt it but many are worried that AI will only benefit the few at the expense of the many. We would be worse off if AI creates a massive “useless class” unable to find a job, but enables a selected few to get rich beyond belief. Answers to this problem aren’t yet clear in the same way that the problem of growing inequality has yet to be tackled effectively. At the same time, AI may lead to higher productivity, potentially leading to cheaper goods for society and overall to a lower cost of life. It could also help us on global matters such as climate change through optimization – something AI is good at. We could do more with less.
What to think of AI
Reading this, it may look like AI has less benefits that drawbacks. It is hard to say what will happen in the future as regards AI. One think is certain, we better get it right. AI has the potential to make vast improvements to the world if it is correctly. It has also the potential to bring about dire societal problems.
In order to get AI right, the first step is to know more about it. This isn’t just something for AI-researchers to do. All of society has a stake in the future of AI. It can be harnessed as a power tool given the right approach, restrictions and adaptation. To be able to set-up those restrictions people need to be aware of what is at stake. To be able to adapt to AI, one needs to understand it: how it works and what it will do. It is safe to say you will have to be affected by the progress of artificial intelligence in one way or another. The more you know about it, the higher the chance that you can use it to your advantage and make sure it does not get out of hand. The less you know about it, the higher the chance that you will be overwhelmed by AI at some point and that it will cost you something – a job for example.
In case you found this article useful or interesting, you can learn more about AI in the next Le M. It will feature a brief about Industry 4.0 to give you a more practical example of AI’s impact on the job market.