AI Learning

Hello Wolfram, I'd like to talk to you today about Artificial Intelligence in the training and development of continuous development. But let's start from the beginning: What is artificial Intelligence explained quite simply?

Wolfram: Artificial Intelligence is a term which is currently very strongly hyped. And as it is always with hypes, not all people have the same understanding of it.

In general, the term artificial intelligence is to be understood as intelligence software technologies that have the goal to imitate human behavior. This can mean, for example, that a robot can walk, grab, or speak.

AI includes the term machine learning. That brings you a little closer to the subject of learning. Machine learning is about software technologies that are "learning" without being explicitly programmed in this way. Or in other words, they do not learn by it, that a programmer gives them rules, they learn these rules themselves. And that by examining data from the past for patterns, and then using those patterns extract it from the data.

AI systems are often described in films and books as human-like systems. But completely without fiction: Still in its infancy, a pubescent teenager or already a senior manager - How far is Artificial Intelligence in its development?

Wolfram: Not as far as it is portrayed in these films, but further than the critics says. A modern computer does not have human intelligence - and that is something that is not possible with today's technology. But I believe also that it's not about that at all. With AI it's always about supporting people, not to replace them. For example, if we look at the whole issue of speech recognition. We all know Alexa, and that's artificial intelligence, or face recognition, or image recognition, or translations... there are already very many examples where artificial intelligence is already being successfully used to improve the lives of the to simplify things for people.

Let us now come specifically to education and training: How does AI develop the way we learn?

Wolfram: In the area of education and training, learners have to make decisions. For example: Which content is right for me? Which courses should I take? Which career paths are interesting for me? AI helps to make these decisions. More precisely said, whenever it comes to making such decisions quickly, AI can play a role. This is also the case with autonomous driving, should the car accelerate or should it brake? Should the car turn left or should it turn right? AI helps to answer such questions on basis of historical data or the current context. AI is a decision support function. That's why it's important to answer the right questions about what you want to answer with AI.

Can you describe a concrete scenario in education and training, in which AI is already in use?

Wolfram: One example is the topic of recommendations. We’re all familiar with this when we talk about Amazon shopping. Users get a very direct prompt, people who have bought article X, also have bought article Y. So you can also tell the learners, all who have learned content X have been interested in learning content Y. Or, this skill profile has also participated in courses A and B.

In E-Learning Punk we deal with continuous development buzzwords that are on everyone's lips right now. But we don't just want to throw buzzwords around us, but also to critically question them, to build up an understanding and explain connections. How do catchwords such as Big Data and Learning Analytivcs go together with AI?

Wolfram: Big data used to be a hype - like AI actually - but then it flattened out again in public perception. Nevertheless, big data is still as relevant today as it was before for many years. Big data says nothing other than data management has drastically changed. The data volume has grown, the data must be processed faster - frequently in real time - and the data is no longer just structured data, as well as unstructured data. This means that the administration and evaluation of data is subject to different laws today than it was many years ago. The topic of learning analytics deals with how we can measure the success of learning. AI can help with both. For example when we look at "Content Curation". Here AI recognises the learning needs by the skill profile and what someone has learned in the past and can collect content from a wide variety of sources and provide learners very quickly. This means that the learner no longer has to search for learning content itself, but the system can offer automatically an assemble of new lessons from past activities. And this helps of course enormous with regard to learning speed and learning quality.

And what role can AI play in VR learning scenarios? Can these become even more lifelike through artificial intelligence?

Wolfram: I think it's not always the individual, but the interaction of VR, AI, Big Data, Learning Analytics... The real advantage of all these new developments for the learner can be found in the interaction.

Our last E-Learning Punk article was about Game-based Learning. If we now connect this to AI in a very concrete way: Are there any learning games that recognise my personality and my abilities, and adapt them to the remaining content of the game?

Wolfram: One of the biggest challenges we face in the area of training and continuous development is self-motivation. And game-based learning is a good approach, because it is about increasing the fun of learning, or rather, learning "playfully". We like to play games. And why? They are interactive, there are rules, there are goals, there are rewards ... The learner no longer sits passively in front of the computer but can intervene interactively. There is no predetermined learning path, instead the learning path develops dynamically during learning through interaction. And that's the nice thing, that it's no longer so predictable. The system can dynamically adapt to what the learner has done before. How my counterpart reacts to my game avatar is not explicitly programmed in the game. They don’t act according to pre-defined rules, but depending on what happens, he decides what to do next. And then we come back to the subject of machine learning - software systems that learn without being explicitly programmed in this way. Without a software programmer having predicted to the characters how and what they should learn, they learn automatically based on data during the game.

Finally, the question about the people behind the technology, what characteristics will we have to bring with us in the future to complement AI and robots? Or to put it another way, what can AI not do in comparison to humans?

Wolfram: Of course, that's the crucial question now. One area with which the AI has problems is decision making. When an AI system makes a decision today, it is very difficult to understand why it has made that decision. It's a bit like a black box. There is no explanatory component. Another topic is intuition. People make many decisions intuitively, that means without knowing why. This is not practicable for an AI system. Also, AI systems always control only a single domain, never several domains at the same time. As already mentioned at the beginning, human intelligence is exclusively entitled to humans. Things that are not based on mathematically describable patterns or behaviours, such as intuition and impulses, cannot be achieved by machines. But, it is not a matter of saying what AI can't do, it's about saying what AI can do. And it's about using AI sensibly in the systems we have today. And that's where we saw that such systems can already do a lot and can already be used sensibly.

Thank you for the exciting interview Wolfram!

The next issue of E-Learning Punk will be published in the New Year and will deal with the topic "Learning by Quizzing."