Yann LeCun – Power & Limits of Deep Learning

I want to share with you one of the most accessible and informative videos I’ve seen so far on deep learning. What really surprised me is the lower number of views compare to others. Future based on facts might not be as appealing as Science Fiction or alarming videos.

Yann LeCun is Director of AI Research at Facebook, and Silver Professor of Dara Science, Computer Science, Neural Science, and Electrical Engineering at New York University, affiliated with the NYU Center for Data Science, the Courant Institute of Mathematical Science, the Center for Neural Science, and the Electrical and Computer Engineering Department.

Video From Prof. Harari

I will begin my series of articles on Artificial Intelligence by commenting a video from Prof. Yuval Harari. He´s a historian, philosopher and not from the technical world which make it very interesting because he is seeing AI from an external point of view.

The main ideas that attacked my attention in this video are:

  • Ai is all about statistical calculation and probability,
  • Actual result obtain with AI are acceptable and will get better in the future because more and more data will be recording.
  • People will soon think that computer knows better and will start believing that it is always correct.
  • Include human biology into human behaviour prediction is the next future revolution.
  • Conclusion: Computer will soon be able to know what you want, better than you,

The presenter is right in explaining that we are talking about statistical calculation and probability. We are calculating probabilities based on the user’s data history and / or comparison with other people who exhibit the same behavioral characteristics than him. The result is good but far to be perfect as Prof Harari recognized.

I have an issue with his reasoning. He believes AI technology will improve and the errors in predictions that we have nowadays will finally become negligible.  Computer will soon be able to know what you want, better than you without considering that these errors are intrinsic of this technology: using the past to forecast the future and a human is a complex system.

Do you remember the butterfly effect? A very small change in initial conditions had created a significantly different outcome. This is due to the difficulties to describe well enough the system at one time and its sensitive dependence on initial conditions. We are collecting huge amount of weather data and process them through supercomputers but I still don´t know if I can organise a barbecue in 4 days and be sure that it will not rain.

The danger is not that the computer will know better than us. Prof Harari mentioned in his video and I fully agree with him.  The danger is that we start to believe that the computer knows better than us and stop questioning the result of its calculation.


Why I want to write about Artificial Intelligence (AI)

I want to explain why I will post several articles about Artificial intelligence.

Before my friend Marwan ask me to join him in Cadmus, I had basic computer skills. I´m engineer but computer science wasn´t my field. Also, I was so busy, I was listening to the marketing talks around Artificial Intelligence, but never really thought about it. It was a black box.

Everything changed with the success of our plug-in “Today for Outlook”. We decided to create new functions to help people even more: The solution to reduce the number of repetitive or boring tasks or just simply make life easier is to use Artificial Intelligence technology. So, I started to learn about Artificial Intelligence.

My goal was not to become a data scientist. I wanted to understand what is possible and most importantly, the problems and limitations of this technology. That’s what I want to share with you in the coming articles.