, , , ,

How to aim for the top and other secrets for success in science: an interview with Professor Emeritus Erkki Oja


Stream it from here:

Download mp3 from here.

In this episode I have the honour of introducing you to one of the giants of unsupervised machine learning and more generally of mathematics applied to statistical algorithms, Professor Emeritus Erkki Oja. I met Erkki met an artificial intelligence meetup in Helsinki where he was giving a talk and we had this interview recorded shortly afterwards. He was born in 1948 in Helsinki and obtained his PhD under the supervision of Teuvo Kohonen (whom you might know from self-orginizing maps or Kohonen maps) then worked at the Brown University with the Nobel prize winner Leon Cooper. The last years before retirement he was the director of the Centre of Excellence in Computational Inference and chairman of the Academy of Finland’s Research Council for Natural Sciences and Engineering.

In this interview we talk about what he learned from his mentors, which habits routines and tricks helped him to get the best out of himself, how does he combine work and family life and many many other topics. We also touch on scientific questions, in particular we discuss modern day artificial intelligence, including deep learning, and its relationship to cognitive science.

Line up:

time – topic; comments

2:10 A little bit about your work.
– Intelligent systems
– Brain
– How to transfer from brain sciences to ingineering etc.
– Neural modelling in the 1970’s, AI…
– Machine learning
– Neural networks
– Hebbian learning
– Oja’s rule
– PCA implementation through Oja’s rule as an emergent property
– Connection to visual cortex, kitten experiments.
– Natural Image Statistics

8:45 How did you feel about not having any citations of your paper for 10 years.

9:45 Plagiarism in science and how to defend your work.

11:25 Large number of citations: >44000

12:30 Relationship between academy and industry, what’s your role?

14:30 What is the motivation for you to do mathematics?
– Algorithms,
– Solve a problems

15:55 Does undecidability ever cross your work (Turing halting problem, computational complexity etc.)

17:25 What is your prognosis for the success of deep networks and deep learning?

18:40 What is the relationship between artificial intelligence and deep learning?

19:40 Relationship to cognitive science and the symbolic vs. non-symbolic cognition debate.

22:25 What were your dreams when you were 20 years old? Did you already know you wanted to be a scientist?

23:50 What about 30?
– Postdoc in US, very formative years, Brown University.
– The importance of being in a top-level research group.
– One must aim for the top.

26:45 How do you achieve a state of concentration? How to achieve top results?
– Being punctual and regular. Go to work at the same time every day etc.
– It is useless to discuss with people who are not more clever than you.
– The only exception was my dog.

30:05 Diet and exercise?
– Exercise is absolutely necessary. Jogging or skiing. Only lonely sports. Because then you can think.

32:00 What should you do if some paper is clearly wrong? How to write a response?

33:25 Motivation. Do you ever have periods of low motivation? What do you do then?
– Realise that you failed and face it.
– A problem is like a bog stone. You have to find a soft place where you can crack it. Maybe you have to
walk around it to find it.

35:20 What should you do if you have too many projects at the same time?
– It is very important to focus.

37:10 Morning routine?

37:30 Are your daughters academic as well?

37:50 Have you ever taken longer periods of time to isolate yourself from everybody else?
– No. But I would have if I didn’t have the family.

38:20 What is the greatest work-related moment that you have experienced?
– Those moments when you know you will solve something. You are still not sure exactly how, but you know it will be a beautiful result. This has happened maybe five or six times in my life.

39:45 What would you tell the 20-year-old self?
– Find someone who is better than you at what you want to do.

0 replies

Leave a Reply

Want to join the discussion?
Feel free to contribute!

Leave a Reply

Your email address will not be published.