# AI is a great opportunity for our civilisational and economic development, but also an enormous challenge

Author: Norbert Biedrzycki
Published: 2019-01-15
Language: en
URL: https://norbertbiedrzycki.pl/en/ai-great-opportunity-for-civilisational-and-economic-development-but-enormous-challenge/
Summary: BrandsIT Magazyn: AI can turn even the strongest legal systems upside down if the legal world does not realise that the sooner we start rewriting the codes, the better for us all.

My article in BrandsIT Magazyn, 13 January 2019.

[Link](https://brandsit.pl/magazyn-01-2019/#dflip-df_18912/1/) to the magazine. My article is on page 18.

-   <figure><img src="/wp-content/uploads/2019/01/BrandsIT-Norbert-Biedrzycki-AI-prawo-1.png" alt="BrandsIT Norbert Biedrzycki AI and law 1" width="866" height="1149"><figcaption>Page 18</figcaption></figure>

-   <figure><img src="/wp-content/uploads/2019/01/BrandsIT-Norbert-Biedrzycki-AI-prawo-2-875x731.png" alt="BrandsIT Norbert Biedrzycki AI and law 2" width="875" height="731"><figcaption>Page 19</figcaption></figure>

-   <figure><img src="/wp-content/uploads/2019/01/BrandsIT-Norbert-Biedrzycki-AI-prawo-3-862x731.png" alt="BrandsIT Norbert Biedrzycki AI and law 3" width="862" height="731"><figcaption>Page 20</figcaption></figure>

-   <figure><img src="/wp-content/uploads/2019/01/BrandsIT-Norbert-Biedrzycki-AI-prawo-4-871x731.png" alt="BrandsIT Norbert Biedrzycki AI and law 4" width="871" height="731"><figcaption>Page 21</figcaption></figure>

-   <figure><img src="/wp-content/uploads/2019/01/BrandsIT-Norbert-Biedrzycki-AI-prawo-5-856x731.png" alt="BrandsIT Norbert Biedrzycki AI and law 5" width="856" height="731"><figcaption>Page 22</figcaption></figure>

-   <figure><img src="/wp-content/uploads/2019/01/BrandsIT-Norbert-Biedrzycki-AI-prawo-6-867x731.png" alt="BrandsIT Norbert Biedrzycki AI and law 6" width="867" height="731"><figcaption>Page 23</figcaption></figure>

The full article:

It often happens that reality slips away from intellectuals, politicians and economists when it suddenly accelerates, leaving us all behind. I have the impression that lawyers leafing through legal codes are having that experience today. They must surely see that many situations arising from bringing machines, robots, computers and algorithms into our lives are not foreseen by legislation at all.

These are no longer (and will be even less so in future) isolated cases in which a judge or lawyer finds no support in the existing provisions. [Artificial intelligence](https://norbertbiedrzycki.pl/en/a-machine-will-not-hug-you-but-it-may-listen-and-offer-advice/) can turn even the strongest legal systems upside down. And that will happen if the legal world does not realise that the sooner we start rewriting the codes, the better for us all.

In the context of artificial intelligence, these questions are urgent, difficult and often not even defined. I will point to a few areas that are particularly sensitive from the point of view of legislation – legislation that must finally face the fact that it will increasingly be ruling on artificial brains.

One of the more popular views on the impact of artificial intelligence on our lives concerns work. Reports that robots will replace us in factories and offices stir emotions. "They'll fire me soon, because a robot will cost them ten times less than me, won't demand a raise and doesn't need a pension" – such voices may multiply in the era of the heralded Industry 4.0 revolution. The argument that eliminated positions will be replaced by new jobs is, for now, barely audible.

In highly automated South Korea, there are already 437 robots for every 10,000 people employed in manufacturing. Japan has 323 and Germany 232. In Germany, a robot's work in the automotive industry costs an employer 5–8 euros, while a worker in the same position earns 35–40 euros.

Robotisation affects many other fields, such as medicine, marketing, media, IT services and even law. At the international news agency Associated Press, about 20 percent of news items and reports are written by applications. Their quality is no longer inferior to texts written by journalists.

Employees of media houses, where numerical data is crucial (and machines handle such data perfectly), also know that automation will significantly affect employment policy. Indeed, even the legal profession may fear robots at work.

As the British newspaper The Guardian reports, an intelligent algorithm was used in work at the European Court of Human Rights. The program found a range of data in the analysed reports and organised it into patterns. On that basis, it was able to predict the outcome of specific human cases with 79% accuracy. According to the paper, tens of thousands of jobs in the English legal sector will be automated over the next 20 years.

Looking coolly at robotisation and automation, it seems that legislative action on labour markets must be taken exceptionally quickly, and labour law faces many important and difficult decisions. Lawyers in this field say that today's codes are essentially outdated already and cannot keep up with reality.

We have more questions than answers here today. Should labour law regulate the share of jobs in a given industry or company that machines may take? Will employers have the right to reduce human staff without limit in favour of machines? Will efficiency standards and targets be the same for machines and people? Will a machine be able to perform managerial duties? Do I have the right to refuse an instruction at work if I know it was generated by a machine? Who will be responsible for potential damage caused by a machine – the programmer, the head of department, the company owner? Will governments be obliged to decide which industries are protected from excessive robotisation, and how?

I think lawyers already working on Industry 4.0 – industry with an advancing social integration of people and machines – can probably count on overtime in their offices.

Littler Mendelson P.C., one of the world's largest firms specialising in labour law, has a dedicated team dealing with robotics and artificial intelligence. In short, legislative processes in this area will develop dynamically, and the advantage of markets based on cheap labour will melt away as robotisation and automation advance.

Copyright questions will become more complicated as algorithms become more present in the media services market and across the internet.

Creators – writers, musicians, journalists – must prepare for times in which creative machines, able to produce texts, music, screenplays and even paintings and photographs, are a fact. Questions about the status of the work and of the creator are increasingly complicated in this context, and the copyright law of many countries is not prepared for the changing situation.

Recently, news spread around the world of Facebook's decision to shut down algorithms that had started using their own language, developed independently of their programmers. This mechanism of self-improvement, using [deep learning](https://norbertbiedrzycki.pl/en/machine-learning-computers-coming-of-age/) techniques that allow computer programs to learn on their own, may have many legal consequences.

If a bot set up to serve customers asking questions is able to create content on its own, who will be responsible for it? If any content generated by machines misleads me, causes lasting damage to my health or harms my business, whom can I sue? How will we treat plagiarism of human works created by intelligent devices? When I buy a work, for example a painting, will I have to confront a blurred concept of originality and doubts about who its author is?

The copyright law of many countries defines authorship in different ways, and the concept of the author – the creator of a message and their role in the creative process, still identified with a human being – is no longer enough. Recently, problems were pointed out with provisions of UK copyright law, arguing that the concept of "computer-generated" content may allow an interpretation in which a work has no human author. It would then be logical to grant rights to the machine. At the time of writing, the European Commission is working on a directive that is meant to resolve the question of legal personality.

US legislation does not define a situation in which copyright would apply to machine-generated works either, but lawyers point out that this issue needs legislative work. The US legal system even lists examples of works that do not qualify for copyright protection because they were created in a "random" creative process, without human input.

Today, most international legislative solutions assume that copyright protects the fruits of intellectual work based on the creative abilities of the human mind. Google recently presented a series of images created by neural networks. You can already listen to music created by artificial intelligence. A well-known record label was long unable to say who held the copyright to such music: the designer of the algorithms or networks, the owner of the server on which the data was processed, or the musician who provided the samples.

But let us imagine a situation in which a machine becomes the "independent" author of a series of paintings that begin to circulate widely. It becomes even more complicated when the works begin to attract buyers or, over time, become the object of manipulation. Will the solution be a provision that the owner of the computer that created the work is its co-author? In Japan, as in many other countries, the solution still in force is one in which copyright does not cover artificial intelligence at all. But the authorities are reportedly working on changes that may see the light of day this year.

Autonomous mechanics, and above all autonomous vehicles on the roads, are another sign of developing artificial intelligence. I follow reports of [Tesla's work](https://norbertbiedrzycki.pl/en/tesla-human-right-error/) and that of the largest car makers with great interest. Because a transport revolution on our roads could become the most tangible example that artificial intelligence is something concrete, not a set of abstractions known only to programmers.

As with labour law and copyright, the legal issues here are becoming increasingly complex. The concept of legal liability for an autonomous vehicle on the road is particularly interesting. It takes little imagination to foresee problems in determining legal liability for possible accidents.

In the event of a collision, danger or even loss of life, who becomes the subject of a dispute, a possible lawsuit or compensation? Who should bear primary responsibility when people die in an accident involving autonomous vehicles? Will the maker of the algorithm that drives the autonomous car be in trouble, the company producing the car, or the driver-owner? And if the latter causes an accident, can they count on adequate insurance solutions that give them peace of mind?

This example clearly shows that technological possibilities are ahead of our times. Practically anything is possible here – whole fleets of autonomous trucks are waiting to hit the road, and the idea of building taxi services on autonomous vehicles occupies the minds of many business people.

It turns out, however, that the main barrier that will long block change on the world's roads is insufficient legal regulation. An autonomous car is not only comfort and – as numerous statistics show – greater road safety compared with human drivers. It also requires serious changes to road traffic and traffic signals. Optimising traffic in the future may also mean integrating engine speeds, video camera images and signals collected by traffic lights at junctions.

Let us imagine a situation in which a deep-learning algorithm statistically analysing traffic at a junction decides that a stretch of road should open a few seconds earlier than usual. And after that decision an accident happens. Who is responsible?

Autonomous vehicles also raise serious dilemmas for local authorities. An autonomous vehicle that moves more safely may also become more efficient. That could lead to a situation in which revenue from renting out city car parks, which may empty, becomes a problem.

The contentious questions around automating road traffic do not concern only cars and trucks. Drones are becoming the machines of our age, and in future so will autonomous ships and aircraft controlled from computer consoles. Add the fact that more and more autonomous transporters will appear in factories to move goods, and the situation becomes even more complex and interesting.

The behaviour of programs and bots capable of shaping public opinion is becoming a serious problem at election time. The ability to provoke protests, predict them, create false information about competing political candidates, steer polls, online hate – all these phenomena also relate to legal regulation.

I believe artificial intelligence in the hands of politicians can be a rather dangerous instrument, and as citizens we have the right to expect certain issues to be codified as quickly as possible.

One of the most important matters affecting every citizen will be privacy. The analytical tools used by banks and insurance companies are increasingly efficient, but how they work may increasingly become the subject of legal analysis.

What data will it be possible to use to analyse the situation of a customer applying for a loan or favourable insurance? Will algorithms be able to combine information such as postcode, skin colour, place of residence and political views "with impunity"?

Will courts in civil or criminal cases be able to rule effectively when the subject of a dispute is a particular behaviour of an algorithm? We still do not know.

The machine as a responsible entity

I have no doubt that one of the most topical legislative issues will be liability for a product and liability for its use.

In situations where the intentionality of actions that interfere with social order becomes relevant, algorithms must become participants in the legal and judicial drama. After all, behind every action of a program there is some human intention. But while I can imagine appropriate provisions and legal solutions that satisfy all parties here, I still find it hard to imagine effective legislation in the face of the challenges that will arise as deep learning spreads – a technology that assumes a high degree of machine agency.

Can such an autonomous machine, capable of deciding to improve its own operation, be treated by societies as an entity with legal responsibility?

I do not want to, and cannot, judge how real the danger is that artificial intelligence will gain agency or become independent of humans. That issue is being debated today by specialists and leading figures in technology-driven business, such as Elon Musk. I wonder whether, given such a potential danger, one legislative solution might be to introduce universal certificates for devices and algorithms put into circulation. But is certifying artificial intelligence logistically possible?

In my view, it is no longer enough to be aware that reality has accelerated and legal codes are ageing at lightning speed. Politicians, legislators and lawyers today bear great responsibility for the changes that can or should be made. If the average citizen is to feel at ease and enjoy contact with sophisticated technology, they should be protected. They should be completely certain that in a court case in which they stand on one side and an algorithm on the other, the human being and their rights will prevail.

Related articles:

[– A machine won't hug you, but... it will listen and give some advice](https://norbertbiedrzycki.pl/en/a-machine-will-not-hug-you-but-it-may-listen-and-offer-advice/)

[– Can machines tell good from evil?](https://norbertbiedrzycki.pl/en/can-machines-tell-right-from-wrong/)

[– Only God can count as fast: the world of quantum computers](https://norbertbiedrzycki.pl/en/only-god-can-count-that-fast-the-world-of-quantum-computing/)

– [Machine learning: computers are no longer infants](https://norbertbiedrzycki.pl/en/machine-learning-computers-coming-of-age/)

– [How machines reason](https://norbertbiedrzycki.pl/en/how-machines-think-2/)
