When we humans hear the word ‘battery’, we use context to distinguish between an energy storage system and an artillery unit. But computer systems think differently. Researchers are now developing a common computer language based on words that are so well defined that they cannot be misunderstood. Illustration by Knut Gangåssæter/SINTEF

Philosophers and technologists join forces to develop a common computer language that can crack some unsolved mysteries

Linguistic logic rooted in classical philosophy can help us in our search for future eco-friendly materials.

  • How do we select materials that enable clothes, furniture and sports equipment to last longer?
  • How should we process raw materials to make it easier to recycle the ‘building blocks’ contained in packaging and white goods?
  • And what environmentally-friendly compounds can we use in our batteries that will extend the range of your next electric car?

Researchers believe that it will now be easier to get answers to these and other questions that are key to deciding whether we humans can succeed in achieving the green transition

Easier, because a team of European research scientists, including some of us here at SINTEF, are in the process of developing a common language for our computer systems. In short, a so-called ‘definition system’, with its roots in classical Greek philosophy, is currently being developed jointly by a team of scientists and modern philosophers.

Goodbye to sources of error

The background to this project is that in the fields of research and industry, computers currently do ‘almost everything’. Our results and documentation are thus primarily in digital form, although for the most part we need people to interpret these data. 

One alternative has been to connect systems one-on-one and use tailored ‘language courses’ to enable them to communicate. This is a major undertaking. A bit like getting Norwegians and Finns to work together by teaching the Norwegians Finnish and the Finns Norwegian.

If information from one computer system is to be used by another system, some form of manual feeding is usually required. This is an ineffective process with many potential sources of error. For example, some years ago, the Norwegian Labour and Welfare Administration (NAV) experienced that this approach resulted in long case processing times.

One alternative has been to connect systems one-on-one and use tailored ‘language courses’, depending on the needs of the respective systems, to enable them to communicate each time a program has to exchange data. 

This is a major undertaking. A bit like getting Norwegians and Finns to work together by teaching the Norwegians Finnish and the Finns Norwegian. But this doesn’t help much if the next day they have to communicate with people from Turkey or Portugal.

The same ‘alphabet and grammar’

But what if we provided all computer systems with a language with common features, just like in our example about the Norwegians and Finns, sharing the same alphabet and basic rules of grammar? All we need to do in fact is to get all systems to communicate effectively.

We are currently developing such a language as part of our various projects  integrated with the EU’s Framework Programmes for Research and Innovation.

“Almost 950 years ago, the philosopher Anselm of Canterbury applied ontology to prove the existence of God. We have not set the bar so high.

The need for a common language is closely linked to the increased use of machine learning, which is one of the disciplines applied in the growing field of artificial intelligence (AI).

Knowledge gleaned from classical philosophy as to how the world can be described is now being applied to benefit modern science, in the form of a common language for computers.
Currently, if six systems are required to communicate with each other (left, above), we need to complete 15 separate ’translation tasks’ (represented by the connecting lines).
If a seventh system is added, the number of tasks increases to 21.
But by enabling the systems to communicate using a common language (the blue symbol on the right), communication becomes much easier.
Only a single translation task is required, and effective tools are now being developed to make this task even simpler. Illustration by Knut Gangåssæter/SINTEF

Describing, communicating and recycling data

Our aim is to develop a formal language that can be used to describe, communicate and recycle data.

The language is ‘formal’ in the sense that it enables information derived from any given discipline to be understood and expressed by machines and algorithms. It can be adapted for use in AI applications, which one day may offer opportunities for the design of the ‘perfect product’ and the revelation of previously unsuspected associations.

It is a ‘language’ in the sense that the information can also be understood and expressed by humans.

Not like us

If computers are to communicate with each other they must use the same words. Such words must be so well defined that there is no possibility of any misunderstanding.

Computer systems are not intelligent in the same way that we are.

But computer systems are not intelligent in the same way that we are. If you hear the word “battery”, you use context to distinguish between an energy storage system and an artillery unit.

If a computer is to be made aware of context, the context must be specified, and formal specifications of this type are called ontologies.

Taking a universe apart…

The term ontology was originally associated with a classical philosophical discipline that addressed what actually constituted reality.

It was used to address questions such as: do chairs really exist? Or are they just an assemblage of particles sharing a special relationship with each other in time and space that simply generates an impression of what we choose to call a chair?

Our projects circumvent this philosophical approach, focusing instead on ontologies’ rules of definition as a means of describing what we experience.

…and putting it back together again

If you look up the word ‘chair’ in a dictionary you will find ‘an item of furniture for sitting on’. Such definitions are also included in the data ‘glossaries’ that we are developing. But we also expand these definitions by breaking the terms down into what we might call their component ‘lego bricks’.

We create definitions for the assumed smallest building blocks that any given ‘thing’ is made of, and then describe how the ‘lego bricks’ can be assembled to make larger objects.

We create definitions for the assumed smallest building blocks that any given ‘thing’ is made of, and then describe how the ‘lego bricks’ can be assembled to make larger objects.

This approach provides us with definitions of how materials and objects are constructed, of the processes that can change them and the purpose of their design. In this way, we reassemble a universe that we had originally pulled apart.

Helping out AI

In the 1980s, researchers investigating AI started to use ontologies to construct databases of terms and the relationships between them. Such databases enabled AI systems to derive new knowledge by means of logical reasoning.

Later, ontologies were used to make what we call semantic networks. This process involved an expansion of the internet that enabled not only humans, but also computers, to benefit from the information.

Then came the turn of the field of biomedicine. Here, ontologies contributed towards our knowledge that the complete human genome was able to reveal previously hidden associations, including those linked to the functions of different proteins. This in turn led to rapid acceleration in the field of medical research.

A ‘glossary’ for the natural sciences

However, the introduction of ontologies to the field of materials technology has been slow because physics and chemistry are such complex and diverse fields.

This is the point of departure for our work. We have initiated ambitious projects involving the construction of a logically-anchored ontology for the natural sciences. This effectively entails a ‘glossary’ that combines elements from widely differing disciplines and fields of research in its description of subject matter, and has been given the name Elemental Multiperspective Material Ontology.

‘Green’ batteries

The development of eco-friendly batteries is a field in which our ontologies can make a difference.

“The development of eco-friendly batteries is a field in which our ontologies can make a difference.

For this we need a language that first explains that batteries are made up of an anode, a cathode, an electrolyte and other components. We also have to define exactly what an anode is. We have to specify its properties, describe the properties themselves and how the anode is related to the other components of the battery.

In order to do this, we intend to use a separate language for the discipline of battery development, based on the basic language that we have previously constructed.

In this way, computer systems used in battery development will rapidly and inexpensively be able to utilise information derived from other technical fields such as relevant results from experiments with x-ray diffraction – a method that provides information about the distribution of electrons in molecules.

No human intervention

All data exchange involved in this process will thus take place without any costly or time-consuming human intervention.

Almost 950 years ago, the philosopher Anselm of Canterbury applied ontology to prove the existence of God. We have not set the bar so high, but we are immodest enough to claim that we shall make a contribution to technological progress by applying classical philosophy in innovative ways.

This feature article was first published in the newspaper Dagsavisen on 26 November 2022 and is reproduced here with the permission of the paper.

SINTEF has contributed to the new ‘common language for computer systems’, as a participant in the following research projects, many of which are part of the EU’s Framework Programmes for Research and Innovation:

MarketPlaceOntoTransBIG-MAPOntoCommonsOpenModelDome4.0VIPCOAT, SFI Manufacturing and SFI PhysMet.