3473 Views |  15

Dr. Claudia Schulz Talks To Durtti About Teaching Computers To Argue And Reason Like Humans

Never pick an argument with the award winning Dr. Claudia Schulz! At least that would be Durtti’s advice. Her extensive AI research has unlocked many of the behavioural cognitive mysteries of human reasoning. Durtti wanted to (politely!) ask Claudia how getting a computer to argue and reason like us will ultimately impact our lives.

What are your current research interests in your role as a Postdoctoral Researcher at TU Darmstadt, Claudia?

People argue and reason about decisions to make or problems to solve, both in their professional and personal lives. I work on tools that, on the one hand, evaluate how “well” people argue and reason, and, on the other hand, support humans in improving their argumentation and reasoning skills.

Remember back in school, when you had to write essays assessing and interpreting biological, historical, or geographical information? My research aims to automatically evaluate such essays and provide detailed feedback.

For example, by indicating that a conclusion drawn by the student was not supported by enough evidence, that too much information irrelevant to a conclusion was given, or that a conclusion simply does not follow from the alleged evidence.

Another exciting direction of my research is the evaluation of large debates. Wouldn’t it have been useful to point out inconsistencies in arguments given for and against Brexit? Or, even better, to automatically assess which side had the better arguments?!
Stay tuned….

You are also a Subject Matter Specialist for U.K. based Outsmart Insights. What does that role involve?

Staying up-to-date with the latest developments in AI, critically evaluating them concerning their potential as well as their flaws, and presenting these findings to clients.

When you were Chair for Imperial College ACM Student Chapter London, you were invited to the prestigious Heidelberg Laureate Forum. What did you most learn from attending the forum?

The empowering feeling of being surrounded by people who are excited about science and technology!

I returned from the Forum feeling extremely motivated and inspired, full of new ideas for my own research.

I can highly recommend interdisciplinary meetings and conferences in general – they make you think about your work in a broader context and can lead to great ideas inspired by research in other areas, which you would never have when attending a discipline-specific event.

How do you think AI will realistically impact society in the next 3-5 years?

This is a tough one, since AI breakthroughs have been happening so fast in recent years.

AI already affects us in ways most people don’t even notice – for example, the automatic correction and suggestion of words when typing a message on a phone, the detection of friends in photos uploaded to Facebook, or the shopping recommendations you get on Amazon.

If I have to make a prediction, I believe that self-driving cars will have a great impact in the next few years.

I also think that we will see increasing use of AI systems that provide support in different professions and education.

The Artificial Intelligence Group

As a human, how do you try to cope with seemingly insurmountable challenges if and when they occur?

I take a break, do some sports, and enjoy the great outdoors!

If you are stuck with a problem you can’t solve, I find the best thing to do is to gain some distance and immerse myself in a completely different activity and environment.

The best ideas and solutions always come to me while I am not working.

You are also a very well established, award winning Conference and Workshop speaker. How (if at all) do you think your presentation style has changed since you first started presenting to audiences?

Most importantly, I don’t consider talks as a way to show how smart you are by making the presentation so technical that only a few people in the room understand it. Rather, talks are an opportunity to share your ideas with as many people as possible.

I think of talks as lectures – with learning outcomes for my audience.

What’s the message I want the audience to take home? How much detail do I need to give so that they understand what’s novel about my work? How does the topic of the talk fit into the big picture and other topics the audience is familiar with?

Furthermore, for me, a good balance of preparation and spontaneity is key to convey both confidence and enthusiasm.

I like to have an idea of what I am roughly going to say, but not to the point where I know every word by heart.

How would you ideally like your career in AI to evolve?

I am quite open-minded about this, even though right now I see myself more as a researcher than a developer.

I currently love academia as it gives me the opportunity to combine my passion for research and teaching, so I could definitely imagine climbing the academic ladder. But if an exciting opportunity in industry or R&D comes up, I would not be opposed to that either.

Finally, Claudia, one of many awards you have won is for Best Graduate Teaching Assistant. With the benefit of your knowledge and your experience as both a teacher and a speaker, what are the 3 most important things you would share about AI with a curious class of primary school children – and how would you go about presenting it?

I would first ask them if they knew anything about AI – no good in telling them something they know already.

My guess is that most children associate AI with robots. So I would tell them that AI is not just about robots, but that any machine or computer program that has a ‘human skill’ is called “AI”. And that there are some things a computer can do a lot better than humans, such as playing chess or finding the shortest way to school, but that, for example, inventing a joke or understanding what is going on in the Harry Potter books is currently very difficult for a computer.

Maybe I would even teach them about the Turing Test and do a live demo with some AI systems, so the children can see for themselves whether or not the systems manage to fool them.

This would lead to my second point; that compared to humans, robots are still not very intelligent. In fact, they are way less intelligent than shown in most movies (I assume that many children will know about AI and robots from movies).

For example, robots cannot explain why they make a decision or why they do something one way rather than another, whereas humans can.

If you see a picture with a cat in it, you can explain why you think the animal is a cat – a robot cannot. Furthermore (and contrary to movies), most AI systems can only perform a specific task, they can either translate languages or cook a meal, but not both. In contrast, humans have all these skills.

Lastly, I would tell them that AI is not just about sitting in front of a computer and programming, or about screwing parts together to build a robot, but that there are many other types of work involved in AI.

Some people try to understand how humans learn the difference between a cat and a dog, how humans make friends, and how the brain works, with the aim to teach robots how to do these things.

There are also lots of fun applications of AI, for example robots that can paint or cook, or that can play games with you. Furthermore, robots can take care of elderly people, and AI programs help doctors to treat patients and judges to make the fair decisions in a trial.

So if you work in AI, there are many different opportunities to do cool things that help others.

Read more about Claudia and her research here.

You can follow Claudia here on: Durtti and LinkedIn.

Claudia is a member of The Artificial Intelligence Group on LinkedIn.