Etnetera group

people
culture

Machine learning: Interview With Anna & Centi

Work hard, play hard – after a proper day of work you deserve some time blow off some steam -or perhaps to learn something new with a Machine Learning course.

Work hard, play hard – after a proper day of work you deserve some time blow off some steam -or perhaps to learn something new with a Machine Learning course.

Anna: Firstly, we got interested in the big data in our innovation track with Štepán Míko and Centi and we wanted to move a bit further in that, but we did not really know how. Centi found the Machine Learning Prague conference and its content sounded pretty good. But I was afraid that I would not understand anything and it would be just a waste of money. The course by Stanford felt like a perfect way of preparation for the conference itself.

How many people joined you, Centi and Štěpán after your email invitation? 

Anna: There were three of us in the innovation track and after the email appeal seven more people joined in. A lot of people asked if they even stand a chance of passing the course and its true that without a certain mathematic background it would not be possible.

Were there any first-timers among you or were you all “Coursera experts”?

Centi: For me it was a first time taking a course like this.

Anna: I have started a few online courses like Android or Big Data but I have never finished any of them. I always gave up after about three weeks.

How did you cooperate during the course itself? What do you see as the main advantage about not going through the course just by yourself? 

Anna: For me the biggest advantage was definitely that I would feel bad about not finishing the course work every week. The guys have often told me that they have done their work on Wednesday already while I only worked on it on the weekends. I also had rough two or three weeks when I had a lot of work on different projects at a time and I spend half my weekends doing homework for the course. Especially Sunday evenings were a bit hectic. If it was not a group thing, I would have probably given up eventually.

ucime-se-za-letu

Dear passengers, fasten your seatbelts and get your highlighters ready. The lesson at 10 000 feet above the ground is just about to start. 

Centi: This is exactly one of the huge advantages. On the other hand, it would be nice if our group worked together physically as well. If every one of us from the group would do something and then we would do the homework together, we would probably learn even more. I took it as learning something during the course so I could do the homework and deliver it but I am not sure if I am not going to forget all that I learned. I guess the the time will show what the course actually gave me.

You used Slack for online communication. What specifically did you usually used it for? 

Centi: We used to send each other certain Windows-based technical issues on Slack but we did not really talk about the topics of the course there. The ethic code of Coursera actually strictly prohibits sharing the homework results with other participants.

Anna: Martin Holečko shared his math vocabulary on Slack, others shared manuals on how to approach the tasks. At first we thought about meeting offline for about half an hour and refreshing our math knowledge together but then we bailed on that – since you only need the basic of linear algebra for the course.

obed

Offline tutoring did not really happen but getting lunch together did. Part of the “study group”, from the left: Štěpán Mík, Václav Bittner, Martin Holečko, Anna Galkina a Ivan Čentéš.

Centi: I was worried about that a bit at first. I graduated from the University of Life Sciences, where math was not that emphasized but the course itself was not really about math that much. We did not really need any paper formulas to understand how the process itself works.

 

Etnetera will pay the entrance fee for a selected international conference focused on machine learning for every participant that finishes the course.” Was this an extra motivation to study for you or was it just a bit of extra something, sort of an icing on the cake? 

Centi: For me the peer pressure of the group was a bigger motivation than the conference. In the end I only participated in the conference for just one day.

What did the course give you? Was it special in any way?

Anna: You could see the level and the experiences of its host, Andrew Ng, who is also the founder of Coursera. He had his lessons well prepared, especially compared to the previous courses that I took and it definitely was not boring. Ng has always tried to keep our attention, he knew how to talk and he mixed up the lessons with interesting side facts.

Centi: The other courses were organized in a similar manner, with combination of videos and quizzes?

Anna: Yes, similarly. But with the Machine Learning course you have the certainty that you are going to be capable of doing the homework because they tell you all you need to know and that definitely does not apply to every course. I have only got stuck a few times so I the quality of the material was definitely on a high level.

What are you going to use your newly gained experience for? Aren’t you going to ditch us now and go work for Google? 

Centi: For me it was a great experience but I cannot really imagine going any further in this topic.

Anička: It was kind of funny that the lecturer mentioned that we now know already more than half of the world’s developers…

Centi: … but the videos were like five years old.

Anna: I have shared an article on Slack later on that presents sort of a learning path for someone who really wants to go deeper and work in this field. It is good to realize that it is a like two-year-task, since there is so much that you have to learn. You have to keep studying and that is very time consuming, the course is really just an introduction. After you take the course, terms like neuron network is not just a vague buzzword for you anymore but you have a specific idea about the diagram, about how to process goes there and back again.

Centi: It feels like jobs in this field can be separated into two different areas. If we would like to take up the work on the AI, we can work with the data and the tools that already exist. Algorithms have been already written, we have specific tools that we just put the data into, something will come out and just interpret that, which makes it much easier. And then there is the second part about writing a new algorithm that I could not see myself working in.

Can you transfer the knowledge you gained into real life? How is AI going to influence your work field? 

Centi: That is partly already happening. Like our Soyka personalization tool which is a recommendation system that also has to learn in a certain way how to offer people what they want. We use it for O2’s Extra advantages for example. That is the most primitive form of learning.

Anna: And we had a similar, a very simplified system in the course: According to your movie ratings, you are given others to choose from. You could have seen a bit more of how the whole thing actually works even when you do not have that much data.

Centi: I think we also talked about automatic categorization of movies based on their descriptions which is also one of the applications of machine learning – content-based text recognition. According to the movie description the app will differ between a TV show and a comedy movie, for example. But you can go much further: Martin showed us Amy The Assistant, which is an AI setting up your meetings. You just give her your email and she takes care of it.

You share your calendar with artificial intelligence and teach her. For example you tell her: Do not plan anything for me on Friday afternoons if the weather is gonna be nice” or “Don’t plan anything on my wife’s birthday”. And then we will just all go lie on the beach. 

Anna Galkina

Anna: You have to sign up and wait to get an account set up. According to one of the lucky ones who is already using Amy, you cannot tell that it is an AI at all. Imagine that AI is setting up meetings based on your calendar and preferences. You share your calendar with artificial intelligence and teach her. For example, you tell her: “Do not plan anything for me on Friday afternoons if the weather is gonna be nice” or “Don’t plan anything on my wife’s birthday”. The author of this idea used to work in big corporations and he has an assistant as every person in the upper management since those people spend most of their time on meetings. Which makes planning pretty much a question of your priorities – how important is the other person or the thing that you need to deal with. And calendars keep changing which is a problem for normal assistants so he tried to replace the human intelligence with the artificial one. It is a nice idea but kind of a sci-fi. The AI will learn for a while and then we will just all go lie on the beach.

Let’s go back to Coursera and online courses. Isn’t classic university education going to become a thing of the past? If you can take up a course at Stanford while having a full-time job, what value has a college education from the past millennium? 

Centi: I definitely see a value in college education because I learn most while learning in a group. And when I study by myself, my motivation to do something more than just to pass the exam is not that big. The old way of teaching makes puts it down a bit but I still see a meaning in taking up a college education. And I see courses more as an addition to what school cannot teach me or a way how to specialize. I hope that online courses will not replace the classic education in the near future. I think that human interaction is essential for the process of learning.

I hope that online courses will not replace the classic education in the near future. I think that human interaction is essential for the process of learning.

Ivan Čentéš

Anička:  I do hope that this kind of education will replace the classic one! I would appreciate if I could choose from who I want to study and to be able to pick from the world’s best and brightest. For me the current way of college education is more of a nice way how to spend five years of your life and have fun than to actually gain some knowledge. When I was in college, it was nice but I lacked the additional value of it. The lecturers told you what you have to do to pass but no one told you what is it actually good for and how we are going to use it or how it is related to what we studied last year. I don’t think that schools of today think enough about that and I consider this system an anachronism from the past. It lacks the fun of learning something new.

 

Centi: I would be really interested about the opinions of the graduates from the American or British universities. If they have a similar view on that or if their system is completely different from ours.

Anna: I cannot wait for the whole system to open up enough so I could educate myself from anywhere in any way and any field I care about and to be able to pick the top sources in the world. My dream is for that to be possible in like ten years from now and for it to work. If that happens than I think that the classic system of five-year-college-education will disappear and a new system will be created, a one offering the possibility to study from whoever I want and from wherever I want. At the same though, I hope that the social interaction aspect will not disappear since that is something you cannot live without.

Centi: Schools should then work as a certain background, since you cannot really study nuclear physics without a lab. And rest could probably work on an online basis – that could be really interesting.

Are you interested more in business or technology?

read it
Other interesting articles