Talk with Yoshua Bengio – Alcove

written by Marie Last updated on November 22, 2017

Published on November 20, 2017

yoshua-bengio-alcove

A few months ago, I was writing that I did not like attending conferences anymore. They were all the same, boring; we’re on our cell phones, we only learn a few useful things. Anyway, not enough to grab our attention on a Saturday when we imagine everything we could do that’s more productive than sitting in a room surrounded by too many people.

Alcove changed that for me. Made me want to attend conferences again. Those organized by alcove are different. Refreshing. It’s a breath of fresh air and lightness that emerges from these events! I loved it!

I was invited to attend the event for free to write about it on my blog later. By cons, everything written in this post is my opinion!

What is alcove?

Alcove is one night a month. The subject of the evening changes every time, but they have already welcomed a large number of public figures, such as Alexandre Taillefer.

Tickets are $30 on average and are limited. As the venue of the event changes every month, the number of seat changes according to the room. It usually is around 30-40 people. The goal is to offer intimate evenings, where you feel privileged to be there.

And it works very well!

 

Conference of Yoshua Bengio

On the 25th of September was Yoshua Bengio’s conference, an expert in deep learning. This is the evening I’m going to talk to you about!

 

Who is Yoshua Bengio?

He is recognized worldwide for his work and research in the field of deep learning, but also to be one of the authors of a famous book in the area.

Mr. Bengio is currently director of the Montreal Institute for Learning Algorithms (MILA). He completed a bachelor’s degree in electrical engineering, a master’s degree in engineering (computer sciences), and a doctorate in computer sciences at McGill. He also did a post-doctorate at MIT!

What is artificial intelligence, deep learning, machine learning?

Artificial intelligence (AI) is a computer program that can do something smart.

Machine learning is a subset of artificial intelligence. It’s the principle that machines can own data, use them, and learn for themselves. DeepMind is an example of Machine Learning. The computer program defeated the Go game world champion by training himself among a significant amount of data containing the game’s expert movements.

When we say “train by itself”, it means that the program performs game movements and analyzes the consequences of these. All these movements are recorded in his data. They add to their knowledge base every time they make a move they’ve never done before … to form a huge database, where the program has had a lot of possibilities. Thanks to his “experience”, he knows what is the best decision to make during a competition.

Deep learning is a subset of machine learning. It involves using machine learning techniques to solve real problems by simulating a network of neurons, mimicking the decision-making process of a human. This technology is used among others by Google for facial image recognition. We will remember the error that occurred at the launch of this feature: black people were identified as gorillas

So… Artifical intelligence > Machine learning > Deep learning! 

The evening

The evening was held at Melisse restaurant on William Street in Montreal. When I arrived, I immediately noticed the atmosphere of tranquility that reigns in this beautiful restaurant. It was a perfect venue!

To relax the atmosphere and get to know our neighbors, all participants had to say a few words to introduce ourselves to everyone before it was Professor Bengio’s turn.

The conference, in the form of a discussion between Frantz Saintellemy and Mr. Bengio, then began. It was evident that the interviewer was very prepared, knew what the teacher was going to answer, and was able to ask questions in a continuous and meaningful sequence.

Around 7:15, we took a break of 10 minutes. We started again around 7:25. Now was the time for us to ask the professor questions. This is the part of the evening that I found the most interesting. The questions of the participants were very concrete and intriguing!

Alcove Yoshua Bengio

Questions from participants

Should we fear artificial intelligence?

His response opened my eyes to this. I think that his arguments are very valid: the fears we have against AI come from movies and science fiction, in which we say that the AI will one day attack us. According to him, we are very far from succeeding in building machines that have an intelligence comparable to that of humans. To be afraid of being attacked, we can compare it to the relationship between humans and animals: we treat animals poorly, but they are less intelligent than us. We are afraid that the AI will become smarter than us and attack us. But, it’s us who design and control the AI. It is inevitable that if we misuse this technology, we could hurt ourselves. But if we are reasonable with it, it can be beneficial rather than harmful!

Why did he choose to stay in Montreal instead of going to work in a Silicon Valley company?

He replied that Montreal is where his life is, and that he would not feel right in California. He is worried about all these companies that are “tearing” Montreal’s talent to the United States. I was shocked by his answer, but I found it good on his part.

He went on to say that companies like Apple are working in a complete secrecy to develop their new technologies. Even people in an office do not know what others are doing to prevent leaks. Secret slows down science. He goes on to say that if Apple does not change its way, they will be screwed. They would have recently realized this since they opened a deep learning laboratory. Within this group, researchers have more freedom of action and publication of their work, but this is not yet ideal.

The same thing is happening in the military right now. They are entirely dropped in AI because they are also in a complete secret. To understand what’s going on, you have to be part of the conversation. If we are in his corner and we do not exchange with his friends, we will always be a little late for others. Open source is essential to stay on the cutting edge of technology!

Alcove Yoshua Bengio Deep Learning

What excites him in AI?

Medicine! He thinks it will not be easy, that there are a lot of barriers because organizations have a lot of concerns about confidentiality. On the other hand, he is anxious that it is possible to recognize cancer cells from x-ray photos!

What does he think about AI and advertising?

It worries him since it’s a way of manipulating the brain of consumers. Even if we think that the targeted advertising that we see does not affect us, it is not 100% true. It is not for nothing that in Canada we forbid advertising dedicated to children. A significant amount of data on each of us will be accessible to organizations. As technology continues to grow, the ability to influence will become even more worrying.

Advertising in the field of politics is also an issue. Technology that falls into the wrong hands or if we do not give ourselves the right collective rules for it to be well managed it can become problematic.

Many other interesting topics were discussed at the conference by Mr. Yoshua Bengio; I invite you to listen to the full podcast if you want to hear the questions! Discover among other things what he thinks of facial recognition!

Conclusion

Professor Bengio did an excellent job of popularizing the conference. It was so exciting to have access to a researcher who talks about his experience, his research, and his life! Especially since he was doing it in a completely open way as if we were in his living room and we had known each other for ten years. It’s not every day that it happens to us!

I went home with a lot of motivation and a smile on my face!

 

To attend an alcove conference

The year 2017 is over for Alcove, but the next event will take place at the end of March, and this will continue until October 2018. Follow their Facebook page not to miss anything!

 

References & credits

Pictures in color are from Mikael Theimer, and pictures from the conference are from Darwin Doleyres.

Sources for the explanation of artificial intelligence, deep learning, and machine learning: Deeplearning4j and TechRepublic

You may also like

Leave a Comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.