Helsinki, Finland – An audience of international peace brokers have gathered inside a room in the historic House of Estates. They have come from South Sudan, the Central African Republic, Ukraine, Colombia and elsewhere to hear a scientist speak.
That scientist is Timo Honkela, and his keynote speech on the second day of April’s National Dialogues conference is titled Peace from a Different Perspective – a Dialogue of a Million People.
It’s an intriguing topic, particularly from a specialist in artificial intelligence.
But 54-year-old Honkela is working on a machine that he hopes will facilitate world peace.
“World peace would be a good goal to work for in my remaining days,” he says, smiling over a cup of coffee during a break in the conference.
His humour is as dark as his coffee. Honkela has been diagnosed with terminal brain cancer, so he has little time left to solve an intractable problem that has plagued humankind for millennia.
Living for a cause
Memory loss appears to be one of the symptoms of Honkela’s cancer and he’s starting to have a hard time remembering words. But he refuses to be anxious about anything – his new handicaps, the immensity of the task before him, even his death.
“Of course, I wouldn’t recommend anyone get cancer to learn how to lead a stress-free life,” he jokes.
Project timetables are normal for research scientists, and Honkela is used to working with five-year project windows. “But in this case I started to think differently,” he says.
“Artificial intelligence and machine learning can be used for something positive within a longer timescale. It becomes realistic; it is realistic even now. It’s not foolish, and it has ground.”
Honkela believes peace – not weapon – technologies should be prioritised and says applications utilising neural networks, big data and digital humanities will be fully at our disposal in some 20 years. Trailblazing advancements, he says, are already under way.
“For example, machine translation is becoming better all the time. Five years ago it was laughable most of the time, but now translations between Indo-European languages are reasonable in many cases. It’ll significantly increase possibilities for human collaboration and communication over the next 10 years.”
Can artificial intelligence help create world peace?
At present, most artificial intelligence technologies are focused on business and marketing applications. We’re using them while browsing through social media, calling service lines and typing text messages with text prediction. So what would a Peace Machine entail?
“It’s a concept,” says Honkela. “It’s not only science, but science-based planning or design of a concept in which the idea is an intention to change the world.
“The assumption is that the vast majority of people would enjoy a full state of peace. Empirically that seems impossible, since wars have been so prevalent. So let’s use scientific and technological means to improve the chance to have peace.”
Pekka Haavisto, the president of the European Instute of Peace and a Finnish parliamentarian with a background in mediation, is impressed by Honkela’s idea.
“… Machines and artificial intelligence can’t substitute human beings, but they can provide knowledge, possibilities and support for peace processes,” he says. “Those processes are often about understanding the language, culture and marginalisation.”
Building the Peace Machine
... Machines and artificial intelligence can't substitute human beings, but they can provide knowledge, possibilities and support for peace processes. Those processes are often about understanding the language, culture and marginalisation.
The first iteration of the Peace Machine will be a book about how artificial intelligence can help to solve human disputes.
Honkela hopes its content will interest the public and that the Finnish original will be translated, at least into English, Spanish, Arabic and Russian.
His approach is multidisciplinary, tapping computer science, linguistics, language philosophy, psychology, sociology, cognitive science and other disciplines.
“When I got my cancer,” he explains, “I was thinking that I’ve been studying machine learning and artificial intelligence for more than 30 years – what could I do to help us in a significant way?
“My objective of world peace sounded, at first, overly ambitious. Humans have been fighting aggressively with each other for thousands of years, so how could it be any different in the future?”
Then he asked his colleagues, whom he describes as “no-nonsense people”, and says they thought the Peace Machine was “a great idea”.
Jorg Tiedemann, a professor of language technology at the University of Helsinki, says: “The first impression when hearing the project’s title is probably similar for many. It sounds naive and over-idealistic. This was the same for me.
“The term ‘machine’ may also lead to confusion, giving the idea that there will be one specific machine that will create peace. When I talked to Timo about it for the first time, I definitely wanted to know more about his thoughts and ideas behind it.”
Honkela forged ahead with his idea. He now has 180 pages and counting.
A crowd-funding campaign secured him transcription assistance – an essential resource, since he lost about half of his eyesight following a brain operation.
Reading and writing have become difficult and he rarely works more than four hours a day at the University of Helsinki. After hours, he sits at home dictating Peace Machine material into a handheld recorder.
Essential to the advancement of the Peace Machine is that computer systems are already starting to be able to analyse the content of human text – in other words, read books. There are some 130 million books in the world, and Honkela imagines how the wisdom and understanding locked in their pages could be utilised if collected and interpreted by computers. With help from machines, cultural barriers that hamper mutual understanding could be significantly diminished, he believes.
“Machines will be able to simulate human behaviour and human cognitive functioning. We don’t need to programme machines through our expertise. Instead, machines can learn in a human-like way to become experts in many fields, thanks to simulation learning,” Honkela explains.
What this means is that engineers would not feed the contents of books into the system the way they interpret them, but allow the computer to study and interpret the words directly.
Tiedemann focuses on machine translation and believes translation technology could drastically reduce language-related discrimination.
“Language technology is essential in any concept that concentrates on communication and understanding. The latest developments in language technology with deep learning as the backbone offer new possibilities for natural language processing with better abstractions that come closer to real understanding …,” Tiedemann explains.
“Conversational agents for natural interaction with machines are not far away and to some extent [are] already reality. Using that power for reducing misunderstandings and miscommunication is one of the goals of the Peace Machine.”
Honkela’s favourite specimen to illustrate the fluid subjectivity of human language interpretation is the word “fair”.
“If we say, for example, ‘This is not fair,’ we think it’s a fact. But actually, what is being meant by ‘fair’? If we take someone in Dubai and someone in Helsinki, or someone in New York and someone in some other place in the world – those people saying ‘fair,’ what they actually mean by that can be quite different,” he explains.
“We need deep linguistic cognitive and computational work in order to actually reach the meanings of the words that people use. That includes the fact that not only do we have problems in reaching each other in one language, but of course we have some 7,000 or more languages in the world.”
Getting humans on board
Yes, people would have to surrender privacy to allow computers to follow their lives and learn their habits, but Honkela isn’t too worried about any potential negative consequences.
“The main concern is that some people start to use machines in a way where machines can take over other people, but not that individual machines could take over all of us people,” he says.
“It’s about how we as humans relate to each other when we have ever-increasing capacities provided by machines. I’d like to be optimistic that we actually have control over these things and suggest that the majority of the people in the world are good people.”
Honkela imagines that one ambitious application of the Peace Machine could be to improve democracy through artificial intelligence. Decision-making by a select few, even when those few are elected, leaves the wisdom of the masses as an untapped resource, and Honkela believes greater participation could help to stabilise democratic systems.
“I would suggest that democracy can be strengthened so that all the people can participate in decision-making. We don’t need to choose one or a few leaders, but we can have all the people be involved, if we so wish,” he says.
“Of course it’s an investment of time and energy, so people wouldn’t need to be politically active all the time, but in the long run one scenario is that we don’t need one leader or parties – we are all going to be Mahmuds, Jameses and Timos discussing with each other.”
The power of emotions
One emerging subject of artificial intelligence research is emotion, because people often make decisions based on feelings and not just facts. In computer science terms, emotions are our main programme.
“When we make big decisions or when we need to react quickly, we go through emotions, and actually there is nothing necessarily wrong with that. We can’t get rid of emotions. We as human beings are essentially emotional beings, so it’s our essence,” Honkela explains.
“In many cases emotions are rational. The question is what we do based on those emotions, and that’s crucial also in relation to the Peace Machine concept and all these questions about war and peace.
“One of those underlying things is that we may have emotions which are relevant for the context and those emotions which aren’t relevant for the context. The latter ones are dangerous ones, because there might be emotional dynamics which are based on some very old experiences.”
Mikko Patokallio, an analyst with Finnish conflict-resolving NGO Crisis Management Initiative, says Honkela is daring to think big.
“I wouldn’t say that the removal of emotions or cultural and societal lenses is a good thing for peace, but having a tool that can recognise and highlight them and, if necessary, filter them can be very useful,” he says.
“Misunderstandings and misinterpretations are a very real part of conflict. If the Peace Machine or any other approach can help determine common ground, points of consensus and mutually understood language of agreement, it would be very useful for us in our work to resolve conflict.”
Honkela is very aware of the power of emotional dynamics. When he was eight years old, his mother committed suicide. It was only after his terminal cancer diagnosis, he says, that he managed to free himself from a trauma-based experience of the world.
He believes the rise of populist political movements speaks to the power of emotional dynamics in sociopolitical contexts.
“I find this Peace Machine concept to be potentially very relevant. We become aware of our own and each other’s emotions, and we can say: ‘OK, we are going in a direction where we’ll start to fight each other.’ But then we can start to see that actually we do that for bad reasons, and through dynamics and processes that are not necessary,” he says.
In late April, a magnetic resonance imaging of Honkela’s brain offered encouraging results – both for him and his Peace Machine. It identified no new active tumours, so Honkela will be on the job for at least another month. But what will happen after that is anyone’s guess.
“I have to accept my own imperfection as a researcher,” he concludes. “I could polish the Peace Machine for eternity, but I have to learn to let go.”