Free lecture on “Nuts & Bolts of AI” with Frank Benda

While exploiting the power of the computer systems, оur own curiosity led us to wonder, “Can a machine think and behave like humans do?  Thus, the development of AI started with the intention of creating similar intelligence in machines that we find and regard high in humans.

Now we’re deep in it, though. There is a vast potential of applications for AI, but it has established a real dominance in fields such as gaming, NLP, expert and vision systems, speech and handwriting recognition, and intelligent robots. So where do we go from here?

This is why we turned to Dr Frank Benda, an AI savant, and a lecturer at the University of Vienna. For a professional of this stature, Frank is amazingly easy to talk to about complex subjects such as AI. That’s why on September 22nd 2020, we had a free Data Science webinar named “Nuts and Bolts of AI” where Frank as a guest lecturer shared his own insights and experiences about the most advanced technology to date.

Brainster: Frank, welcome! We rarely start these interviews traditionally. Let’s cut to the chase, Auto ML – threat or encouragement?

Frank: Encouragement – absolutely. Covering the main part that programmers are working on currently in their daily life and replacing these processes with automated ones to apply machine learning approaches is pretty much what the trending topic digitalization is all about. We shall not try to prevent this process to happen at all but think forward: If we save time and effort within the framework of implementing these approaches to real-world problems where else could we spend this time on instead?

Brainster: On a more serious note, AutoML 2.0 should help in the most challenging parts of the Data Science workflow, such as feature engineering. Which other benefits could we expect with its further development? What are the improvements that you would like to see in your own work process?

Frank: I am very curious about the level of automation that can be achieved. For me personally, it would be interesting to get a detailed insight into the entire implementation phase of such a project. So I’m very excited.

Brainster: Now we can go back to the start. 🙂 Please, introduce yourself to the Brainster Audience. Tell us more about your career. What made you pursue a career in Data Science?

Frank: At the very beginning of my PhD program and employment, there was only a term: “Industry 4.0”. We were given a cyber-physical system of a real-world problem setting of our industrial partner and wanted to solve it with an out of the box approach – instead of the “typical” optimization algorithms that we made use of so far.

It was pretty obvious that we wanted to give machine learning a try within this context because of some major reasons: It is easy-to-implement, easy-to-adapt, and easy-to-apply. This was pretty much my first contact with the broad field of possibilities that arise with machine learning approaches.

From there I published papers and my PhD thesis containing different machine learning and data science approaches based on those real-world applications, gave talks on conferences such as IJCAI, and attended a lot of different workshops to gain more and more know-how of suitable problems that data science can be applied to.

Brainster: On September 22nd 2020, we had a free webinar titled “Nuts and Bolts of AI”. Give us a small overview of the webinar?

Frank: The webinar roughly included the topics that were briefly touched on here: What comes after the hype? How can we set up helpful implementations in business processes that add real value? Why have smaller companies in particular been so critical when it comes to the buzzword “artificial intelligence”? And how can we convince exactly these companies when and that data science can be used appropriately?

Brainster: As a professional, you might be tired of the hype surrounding Data Science, but as we can see that hype is going nowhere. Did you expect this to happen though? How Data Science became the “hottest job” of the 21-st century?

Frank: It was pretty obvious, wasn’t it? As soon as the first marketing agencies promised the robots that will solve all of the companies’ problems with hyper-intelligent, artificial intelligence within their advertisements 😉 Of course, they could not cope with the level of expectation that arose with these hyped not-so-realistic robot-like implementations – as they weren’t realistic robot-like implementations. But some years later, after the first hype cycle is over, it is even more important to determine the reasonable applications for Data Science. We should raise more awareness of in which relation Data Science approaches might really add some value. That is one of the main goals for teachers and lecturers: to gain trust in the industry again. There are so many problem settings to test Data Science on, but with a more grounded point of view instead of exaggerating the expectations with typical marketing phrases.

Brainster: But people who have a strong interest in Data Science are usually discouraged to start studying seriously due to the age-old fear of coding. How do we battle this stigma? In your experience as a lecturer, which part of Data Science is the hardest to master?

Frank: Of course, I already learned the basics of statistics during my studies but I didn’t have any specific coding skills at all.

My introduction to the topic was through the conceptual phase of our real planning problem: What is the problem setting all about? How do I convert this setting into a machine learning approach? What do I actually need for this? First a production simulation. Then I need a mathematical model that solves the training instances. After that, I have to read in these training instances and filter out information in terms of parameters to train my Machine Learning approach. And finally, I also have to performance check my approach.

So I worked my way up to the topic of machine learning bit by bit. In each phase, I improved my basic understanding of the statistics in the background, programming and of course the possibilities of different Machine Learning approaches and which ones are suitable for my problem setting. It was a tough but very interesting road until the final solution.

Brainster: You are a Machine Learning Engineer who is active in both industry and Academy. Research in Data Science is making huge strides even though it’s for the long haul, while in business there are implementations on a day to day basis. On that note, could you describe your typical working day? What are you currently working on?

Frank: After I published machine learning approaches during my PhD studies and as a university assistant, I now dedicate myself to imparting my knowledge and experience to convince others of the fascination and possibilities of this topic.

Brainster: It probably varies depending on the project, but what is your preferred toolkit as an ML Engineer?

Frank: For me, it was the common toolkits, such as Numpy, TensorFlow, and R.

Brainster: COVID-19 increased the pace that the digital transformation is taking the world over with. AI is slowly integrating into all business sectors and is becoming a “staple” technology. With this, the media is slowly pushing the narrative that soon AI will be able to solve and prevent major adversities like natural disasters or pandemics. Do you agree with this statement? If so, how soon is “soon”?

Frank: I have colleagues who create disaster scenarios, optimization models. So I know that these are highly complex constructs with a multitude of parameters. Whether it will be possible soon to sensibly replace these existing models with artificial intelligence and whether these will actually be better than the conventional models is difficult for me to state at the moment. If I doubted it, I might soon be found to be wrong 🙂 So I hope that such disaster simulation-optimization models can be improved even further with the proper use of properly applied artificial intelligence.

Brainster: So, singularity by 2045. Fact or fiction?

Frank: We seek for human enhancement (e.g. Neuralink), Facebook buys CTRL-Labs, and James Lovelock in his book “Novacene: The Coming Age of Hyperintelligence” describes artificial or hyperintelligence not as the downfall of humanity, but its salvation. Is singularity going to happen? We have already made some misjudgments as human beings, but I think the singularity will emerge. This is one of the most interesting topics to be discussed from a more philosophical point of view and definitely with a beer in a cosy bar. 🙂

Brainster: Okay Frank, before we’ll let you sign out, what’s your message to our upcoming students at the Data Science Bootcamp?

Frank: Start simple, improve step by step and don’t be afraid of the huge topic and possibilities of data science. Discover it piece by piece, try and fail – and try again. Use your creativity and enjoy it when your idea, your approach actually works and can be implemented in a real-world setting. It will be YOUR contribution!

Watch the recording of our webinar “NUTS & BOLTS OF AI”, with Dr Frank Benda, as a guest lecturer.

If you have ever thought about a future-proof career in Data Science, have a look at our remote Bootcamps.

What do you want to be when you grow up?