If data is the oil of the 21st century, then AI systems are the engines of the digital revolution…

What we might learn from around 200 years of experience in dealing with the opportunities and risks of engines for dealing with the opportunities and risks of AI systems.

?

A contribution to the current discussion by Thomas Schneider, Dietikon (Switzerland), 14 May 2023

?

The anxiety that can currently be felt in the media and also among many people regarding how to deal with artificial intelligence is based not least on the still widespread idea that AI systems are superhuman-like robots and cyborgs that are more powerful than we are and will dominate us. But the AI that is spreading today in all areas of our everyday life is something else: AI is the digital engine that eats up data (and electricity) and - to use Goethe's Faust - holds the digital world together in its inmost folds.

?

Speaking of engines: there are not only parallels between the economic and social transformation potential of engines in industrialisation and AI systems in digitalisation. There are also parallels in how people have reacted to the two revolutionary innovations. So, in order to find an adequate way of dealing with AI in the present and the future, it is worth looking at how we have dealt with the opportunities and risks of engines over the last two centuries. And how people have tried to develop sets of rules so that engines are used for positive goals as far as possible and damage and catastrophes are avoided.?Since we have only been moderately successful in the latter, we would probably have to try to do a few things better in dealing with AI....

?

With both, engines and AI systems, a long time passed from the invention itself until they became widely accepted. After the invention of the first heat engine, which is attributed to the Greek mathematician and engineer Heron of Alexandria in the 1st century of our era, it took over 1500 years until an environment was created in the course of industrialisation that allowed engines to be used in machines that could also profitably replace human and animal movement with mechanical movement. And until the raw materials for the necessary fuels (coal, oil, gas, electricity) for the engines were available in sufficient quantity and quality. In the case of AI, it did not take quite as long, but it did take several decades until enough data were available as raw materials and sufficient computing capacities were created to use AI productively.

?

From euphoria to ignorance to panic, the contemporary reactions to the spread of engines in machines was similar to the reactions to the spread of AI in computer systems today: In 1832, after the Zurich government had spoken out against a ban on weaving machines, textile home workers and small factory owners set fire to the mechanical spinning and weaving factory in Uster near Zurich; out of fear of losing their jobs. At the beginning of the 20th century, the last German Kaiser Willhelm II thought the automobile was a temporary phenomenon and is often quoted as saying: "The car has no future, I believe in the horse." There were people who chased cars because they loved the taste of the exhaust fumes and the oil. The government of the Swiss canton of Graubünden, on the other hand, banned the driving of automobiles on all roads in the canton in August 1900 – a ban that was only lifted again 25 years later by referendum.

?

New technologies change our world first minimally, but then profoundly

?

The first engines were not very powerful, but they required a lot of maintenance. Over the decades, however, people have developed engines that can do much more than a human or an animal. Engines have been used to drive machines whose capabilities have made life easier for people and have also made long-cherished dreams of mankind come true:

?

Engine-driven machines have been able to massively increase the production of goods and food, thus fighting hunger and generating wealth. We use machines to build huge constructions such as skyscrapers, bridges and dams. We have created means of transport that enable us to move living beings and goods over great distances in a short time: on land, on water and in the air. Some have even managed to fly to the moon! Engines are also used to produce works of art and cultural goods that were not possible before.

?

At the same time, the use of engines continues to cause great damage to people and nature to this day: with the help of engine-driven machines, millions of people and animals were and are controlled, dominated, abused and killed. The two world wars of the 20th century could hardly have unleashed such destructive power without engines. In addition, engines and the energy sources they use continue to have massive impacts on our environment, which we have only managed to control to a limited extent to date and which, with climate change, present us with one of the greatest challenges for the survival of humans and nature.

?

Until a few years ago, AI systems were only able to perform simple and very limited tasks – despite a great deal of effort and just as much data - and usually only in poor quality. However, the systems have rapidly become more powerful and efficient. In certain areas, they have already outperformed humans; in other areas, this is yet to happen. What the biggest positive and negative effects of AI will be is still speculation at the moment.

?

Control over new technologies and the raw materials and energy sources they require influence the balance of power

?

Those states that had better engines and machines, and the raw materials and energy sources needed to run them, were able to prevail militarily over others. Powerful companies have emerged that play a dominant role in the production of engine-powered machines. In the field of raw materials and energy sources, too, a small number of powerful players have emerged who play a dominant role worldwide. Instead of technical knowledge and natural resources being shared among all people, a few players have managed to concentrate the benefits of value creation among themselves, while the damage and losses caused by the use of engines and energy sources have to be borne by all people, especially by the weak and underprivileged.

?

With the ongoing digital transformation of our economies and societies, there is also an increasing concentration of technical knowledge, capital and raw materials (data) among a few players who have achieved dominant positions in various fields worldwide. Some private companies have achieved a value creation that exceeds that of entire countries many times over. If we want everyone in the world to be able to benefit from the opportunities of using AI systems and data, and if we want to achieve the UN Sustainable Development Goals, which call for no human being to be left behind, then we need to think not only about the governance of AI systems, but also ask questions about the ownership of and access to data.

?

Differentiated sets of rules are needed to set the right incentives for the use of complex and interdependent systems

?

Engines are rarely machines on their own, but part of more complex machines that process materials and produce new goods through automated mechanical movement, or that transport living beings or things from one place to another. There are identical motors that are used simultaneously in different machines, for example in cars, in ships or in production machines. Depending on the area of application, the same motor can produce completely different positive or negative effects. Therefore, depending on the area of application, there are also different technical and legal regulations, not only for the engines and machines themselves, but also for the people who develop and use such engines and machines. These regulations are also shaped by the respective cultural characteristics of the societies in which the engines and machines are used.

?

There are different regulations for the production of different types of vehicles (automobiles, ships, airplanes, etc.) and also for the different uses of these vehicles (transport of passengers or goods, sporting activities, research purposes, etc.). There are also regulations on how engines must be matched to the other components of the machines they move (such as, in the case of a car, the chassis, brakes, etc.). There are also different safety or environmental regulations for engine-driven machines that produce and process goods, and for their use by people, depending on their purpose and potential for damage.

?

There are also different requirements for fuels, depending on the engine and the intended use. There are also different regulations for the infrastructures used by these machines, e.g. roads or airways, depending on the risks and damage potential. Also for engines and machines that are not used for locomotion but for processing goods, there are a multitude of regulations and liability rules, depending on the respective risks and damage potential for humans and nature.

?

All these innumerable laws, technical standards and other sets of rules cannot be considered in isolation from each other, but are more or less closely related and interact with each other.

?

In order to ensure compliance with the various regulations in the different areas relating to the development and use of engines and machinery, a large number of authorities and instances have been created over the years that are responsible for individual or several specific aspects. These bodies also interact with each other.?

?

Like engines, AI systems are used in very different contexts with different opportunities and risks. Because the use of one and the same AI system in the production of goods can have different effects than in the media system, the judiciary or in healthcare, differentiated sets of regulations are needed that do justice to the opportunities and risks of an AI system in each area of application. Regulation must also take into account the role an AI system plays in a decision. Similarly, it makes a difference whether it is offered on a market where people can choose or whether it is used in a state or private monopoly.

?

Depending on the context of the application, there will also be different requirements for the data used, the raw materials for AI systems. The same applies to infrastructures used for the use of AI-based services, e.g. digital platforms. Regulatory frameworks adapted to the respective context should enable economic and societal innovation as well as ensure compliance with fundamental societal values and individual freedoms and fundamental rights and meet the requirements of sustainable development. This will require not only a multitude of legal, technical and societal standards, but also a multitude of bodies, each in its own area of competence and all together for the entire digital space, to ensure compliance with these standards. These bodies must interact in such a way that the overarching goal of using AI for the benefit and not for the detriment of humans and nature is achieved.

?

Appropriate international interoperability of regulations is needed

?

Depending on climatic and topographical conditions, but also on economic, historical and cultural characteristics, different societies have developed different sets of rules for the development and use of engines. Thus, not only do different countries have different technical regulations for the same engines and machines. There are also different regulations for the use of infrastructures and also the requirements for the people who use the same engines, machines and infrastructures: There are countries with left-hand traffic and others with right-hand traffic on the roads. Alcohol limits for drivers and safety regulations for production machinery may also vary between countries.

However, the more cross-border a field of application of engines and machines is designed, the more a certain harmonisation or at least interoperability can be observed. For example, the rules in air transport are more harmonised than in road transport.

?

Also for the development and use of AI systems and AI-based services, regionally different sets of rules will probably emerge due to different economic and social traditions in certain areas. However, if one wants to enable cross-border use of AI systems, data and AI-based services, a certain harmonisation or at least interoperability of systems and standards will be necessary.

?

Governance mechanisms must evolve with the technologies

?

If one compares the engines as drivers of the industrial revolution of the 19th and 20th centuries with those of AI systems as drivers of the digital revolution of the 21st century, there are not only many parallels, but also differences. In order to deal adequately with AI systems, their governance must also take these differences into account.

?

While motors replace physical force and movement with automated mechanical movement, AI systems automate cognitive activities. Engines are made of matter, are located in a specific place and cannot be transported to another distant place or used for another purpose without effort. Energy sources such as oil are also material and thus strongly location-bound. AI systems and also data as their raw materials are largely dematerialised and can be moved much more easily from one place and from one application context to another. Building an engine several times requires material and effort. Oil can only be burnt once. AI systems and data as their fuels, on the other hand, can be copied almost at will and used multiple times in different places and contexts. Moreover, AI systems can, much better than engines, learn by themselves, develop themselves further and improve and change their performance and results through training.

?

The development and use of AI systems is therefore much more dynamic and also more unpredictable than that of engines. Rulebooks for AI systems and services that build on them must be able to deal with this characteristic.

?

The industrial revolution has already had an impact on the governance systems of societies. In many countries, national and regional parliaments and party systems have (further) developed, in which local populations, usually based on their economic or social roles (entrepreneurs, workers, farmers, clergy, civil servants, etc.), have formed political interest groups (parties) and fought over and passed laws in geographically composed parliaments.

?

Although the objects to be regulated have become massively more complex and dynamic and, indeed, more global in the last 200 years, the governance structures have hardly changed at all in that time. Legislative processes are still organised nationally and regionally and usually still take years. Often, however, the laws are already outdated by the time they come into force and can only be enforced nationally to a limited extent. So, we are still trying to solve the challenges of the globalised digital world of the 21st century with the nationally oriented governance structures from the 19th century.?

?

In order to be able to develop adequate sets of rules for a highly dynamic, highly interconnected, globalised, digitalised and dematerialised economy and society, we need to make the structures of administration and politics much more agile and dynamic. We do not need to destroy the existing silos, but we need to make them permeable enough so that the experts and stakeholders needed to tackle a complex challenge can come together quickly and contextually - and across party, stakeholder and national boundaries - to first understand a challenge with all its components and interdependencies and then work together across silos and boundaries to find appropriate solutions that can also be implemented within a useful timeframe and work for all stakeholders. In a world where the applications we use go from one beta version to the next and software updates happen almost daily, we have to find ways to further develop the rules and regulations not through total revisions that take years, but through repeated updates at much shorter intervals. To achieve this, the people involved also need to work and think more in dynamic networks - because only with collective human intelligence will it be possible to develop adequate sets of rules for artificial intelligence.

?

Global governance needs global core values

?

We will only succeed in developing an appropriate and functioning set of rules for AI systems as engines of the global digital revolution, if we manage to agree on the basic values and goals underlying this set of rules. We need a broad and inclusive societal dialogue at all levels - national, regional and international - about how we want to shape the digital world of the future and for what and how we want to use the engines and the other components of the digital world. For this, the political structures, from national parliaments and public spheres to international organisations and processes such as those of the Council of Europe, the OECD and especially the various UN organisations, must also be further developed so that we can agree on the most peaceful and sustainable use of new technologies for the benefit of all people in the world. New inclusive multi-stakeholder dialogue platforms such as the UN Internet Governance Forum and its regional and national offshoots can play an important role in identifying challenges and developing shared approaches to solutions. But we also need new more agile and inclusive collective decision-making mechanisms to ensure that humanity as a whole, and not just a few, can exercise control over the engines and fuels of the digital revolution.

?

We are dealing with AI systems as the new engines of the digital world. But above all, we are dealing with people who invent, develop and deploy AI systems. So, we should not be paralysed by fear of inhuman AI systems. Rather, with sobriety and common sense, but above all with respect for all people and nature and on the basis of the fundamental values of the international community, we should pull ourselves together and make a concerted effort to learn from the mistakes we have made in dealing with engines and to create a global set of rules that will allow us to use the opportunities of AI for the benefit of all people and to avoid damage to people and nature. Together we can achieve this!

?

?

Thomas Schneider is currently chairing the Council of Europe’s Committee on Artificial Intelligence (CAI). He has a degree in history and economics. These reflections are his personal thoughts. He used an AI system to translate the original German text into English.?