12 minute read
Artificial intelligence is the main discussion topic around the world. Starting from dystopian sci-fi scenarios ending up with giving a citizenship to a robot. Before getting down to cracking the subject, it is essential to understand what AI really is and how will it change our world. To help us with this difficult issue, Niilo Fredrikson gives us an insight into AI operation details and its development. Bath’s academic Joanna J. Bryson explains societal standpoints and the challenge of anonymity in a digital era. And last but not least, entrepreneur and founder of Headai company Harri Ketamo describes how global education can benefit from AI and what is the most important human skill that will help humanity to stay away from dystopian future.
“AI has already changed our lives fundamentally.“
Niilo Fredrikson – inspired by growth in nature and humankind is an expert with outstanding record and understanding of current and future state of technology. Even as a child, Niilo had an interest in technology and established his first company before university years. He continued his career in Microsoft as a Global Director. Today, as an Executive Vice President of Comptel, Niilo leads Intelligence development making the world an easier place to live.
What is Artificial Intelligence?
One of the definitions is AI meaning computers performing tasks or functions that typically only human would be able to do. However, there are some challenges with this statement. If to look 30 years back in time, a notion of computer winning against a person in chess was considered an AI. A couple of years later, people started saying that it’s just an algorithm, a powerful computer. Technological development is so fast that at times it is hard to grasp. Recognising human facial expressions from a live video stream by machine was considered a science fiction, and today it is a reality. Is that AI? To me, all of those developments are AI, but if you ask someone from the future, they might have a different opinion as technology would be a lot more developed.
When will AI be a norm?
AI is a norm already today. AI as in Artificial General Intelligence is a different term that would be much closer to human brain being able to learn and create by itself. There is nothing even close to it existing today. All the AI of today is more or less domain specific. Will there be General AI available 2030? Or 2100? Nevertheless, it is a sure thing that every year we will see more and more AI technology being deployed. A lot of things we thought only humans could perform will be done by machines. It is a gradual evolution and blends of technologies as well as tasks they are performing.
How AI change the life?
We are already on the wave of this change. In my opinion, AI has already changed our lives fundamentally. Interaction with AI in the future will follow the same dynamics as today. AI is already deployed by many sales departments. The manager receives the recommendation of the offering to the client based on data analysis performed by the software. Google or Amazon have transformed our lives with progressive deep learning. Also, the flights are mostly operated by autopilot, making flying safer.
Google created AI that taught itself to walk. What kind of data is required for AI for self-learning?
It requires very similar data that a human child or toddler needs to start walking. First are a sensory input, understanding of balance, position, space. Also, a role of parents who would encourage first steps and help toddler, as well as give an example. When we talk about machine learning, there are two categories: supervised and unsupervised learning. In supervised learning, there is a confirmation process with a person indicating that the chosen direction is correct. Unsupervised learning is when a machine is being fed with data and finds structures to it without a defined goal. Using this method, new patterns are discovered and established. For instance, by giving a data set of different people and all the gens that they have. It is possible for machine learning system to create a grouping of the people and deliver insights is a non-biased and precise way. One key thing that makes difference form unsupervised is that there is a feedback loop.
I have always been inspired by growth starting from nature, trees, kids, as well as technology and its applications development. I am also very inspired by seeing all of us growing as human beings triggered by a community, conversation, technology, challenges.
“The costs of humanlike robots might outweigh the benefits.“
Driven to explain the nature and character of Artificial Intelligence to the world, Joanna J.Bryson is a researcher of AI ethics at the University of Bath. Her transdisciplinary research on the structure and dynamics of human- and animal-like intelligence allows us to understand the future with AI and robotics, and to define legislative steps needed to be taken for the balanced social-economic environment. Joanna shares with us her views on the current status of AI as well as the societal impact of new technologies.
What is AI?
There isn’t “an AI” for me. Intelligence itself is an ability to adjust your actions to the context, take advantage of opportunities. AI is just intelligence that is being built deliberately by humans. The key word is “artifact”, it means that AI is made by humans, who are then responsible for that artifact being able to make decisions and operates correctly at a time.
John Searle argued that his Chinese Room Argument shows that there is no AI at all. In this thought experiment, a person who doesn’t know Chinese is put in a room with instructions on the walls and there is another Chinese speaking person in the next room. The Chinese speaker sends a question in Chinese to a non-Chinese person in the room, who answers to it using the instructions. He or she understands neither the question nor the answers and nevertheless gives an appropriate and correct answer. John Searl argued that this means there can be no AI, but most researchers think this thought experiment is flawed. First of all, such a system couldn’t be built, but secondly, if it did, then the system as a whole would know Chinese, even if some of its components didn’t.
AI the future or reality?
We already have had intelligent artifacts for a long time. For example, a thermostat can be considered AI. It notices that it is too warm and makes house cooler or the other way around. By that definition, we have had AI for million years. In fact, technology is often considered iAII with only one component of intelligence. We speak of machine vision as AI, even it is not generating actions. Other expert systems given a task may respond with the answer, but any action is taken by the person who gets the information. As a result, we often call things AI even though it just augments our own intelligence.
Obviously, the AI is improving as we have become better at creating intelligence since the digital era. There have been a lot of breakthroughs and changes happening in the last ten years. That fact doesn’t mean the present rate of change will keep going forever. We as a society keep learning more and more. AI has been improving rapidly as we have gotten better at uploading our knowledge to AI, but we’re pretty much through with that. Now and in the future, AI will help us to expand our intelligence threshold – what we know and what we know we can do. The development of AI has been moving fast gaining the level of human intelligence and now it will be slowing down to moving forward at the same pace with humans.
Robot humanization – is it needed?
Humanization may help people to understand the robot, to see how to use it. But I actually think, it makes people mistakenly believe they understand how a robot really works. For example, they might forget that the AI device that is sitting at a dining table and telling the weather, also can record entire conversation and upload it to the cloud. Even if you trust perfectly the company that stores data, you still should remember that data can be leaked or hacked. The costs of humanlike robots might outweigh the benefits. Apple figured out to put one button on the IPad, it made a device a lot more usable and it has nothing to do with being humanlike. We have to be more creative and find some other ways to help people to use an AI that doesn’t involve confusing them further by making them think it’s a person.
For example, to talk about giving a robot citizenship, as was a recent case in Saudi Arabia, everything goes wrong. It’s about having a supposed equal you can turn on and off. How does it affect people if they think you can have a citizen that you can buy.
Privacy and AI
We need transparent and ethical ways of using AI. Part of the problem is that a person loses anonymity. AI can search through and link together a huge amount of data. What used to be a momentary action or statement with no consequence is now recorded. There was no perfect record showing personal political opinions with 30 years of history. Today by analyzing someone’s 126 likes on Facebook, Cambridge professors can tell what would be his or her vote better than his or her partner can. If AI is getting to know you better than you know yourself, that would be a new kind of loss of privacy where one person’s moves can be predicted and he or she can be manipulated. The best solutions I have seen so far is European GDPR which retroactively allows people to claim that an unanticipated use of data is illegal. However, Europe is largely alone in this level of personal data privacy legislation. In the USA there is private data ownership, and China is making data a governmental issue with whole rights to use it.
I have always found those sorts of problems like AI ethics very interesting. Now is the time that I and my peers need to take a turn trying to help the planet. I am fascinated by the shape of the problem and excited to contribute to solving the challenge.
“Critical thinking is the skill that will keep us above the water.“
Harri Ketamo is an expert when it comes to Learning Sciences, Data Mining, and Artificial Intelligence. His excitement of creating a machine that can mimic human learning enabled Harri to establish a prize-winning Headai company that innovates its way to challenge current education systems. Along with successful entrepreneurship career, he is a Senior Fellow at University of Turku and Adjunct Professor at the Tampere University of Technology with more than 80 international peer-reviewed research articles.
Why do we need AI in education?
Naturally, everything can be done without AI if we have enough money and people. Right now, the world is lacking both. There are some 10 to 100 million need for teacher globally if we want to educate everybody. The only solution is to have AI decreasing the cost and educating more people in very different and fragmented areas.
Does AI change the teachers’ role?
A teacher no longer needs to stand in front of the class delivering the content. Information delivery could be digitalized, and teacher can concentrate on each student, providing one-to-one education in a personalized manner.
A school will become more for human connection, not technology. Early years of social development are essential to developing social skills, leadership, empathy. In life university and vocational education, teachers have to challenge the thinking and ideas of their students. The lecturing and content is not a point, but a development of independent thinking is. With AI as a content delivery and teachers as mentors, education will become a lot more personalized and productive.
What is teachers attitude in Finland about working with AI? Are they excited or deny an opportunity to utilize this tool?
More than ten years ago we introduced our first AI that was aimed to help teachers by delivering personalized learning and pointing out students’ strength and weaknesses. That time, teachers felt like it was taking their jobs and they declined the programme. Nowadays, there is an understanding that AI teaching programs are tools supporting teachers and allowing them to concentrate on students. In that sense, climate among teachers in Finland has changed towards accepting digital assistance.
What kind of education would we need when AI becomes our colleagues at work?
Critical thinking is the first. According to many of science fiction, AI plays a role of our enemy. Nevertheless, it’s just a well-programmed application; it’s a tool. But the tool makes mistakes. For instance, using limited data sets will provide limited information to the programme. In some areas, human accuracy is 92% when machines estimation is 96% accurate. By combining those affords we can even reach 98% level. It will depend on human ability of critical thinking and ability to operate with new technology. Moreover, in the informational overload of today, ability to think critically is essential. There are hundreds of news every day, which of them are true.
Downhills of AI
It is worrisome if people start relying too much on AI and slow down own ability of critical thinking and development. If we start to trust too much in decisions made by machines without being able to evaluate their work and the direction of development, we are taking steps[nbsp]back in such domain. Always, no matter what AI is predicting, it is based on previously produced data, and there is no evolvement in that. For instance, in medical science, we can build AI that can recognize cancer more accurately or other diseases. But if we stop educating doctors, stop medical research, it takes only a decade for the medical system to be completely broken.
I want to do things nobody has done before, and discover undiscovered. The main thing is to follow own passions in life and do things I truly enjoy.