AI Revolution – Transcript

0’21’’Quentin Bohlen, Journalist

Artificial Intelligence allows machines to perform tasks and solve problems that were previously reserved for humans.

0'26'' But above all, it can also learn by analyzing the millions of data available thanks to digital technology and the internet.

0'35'' The scope of  application of Artificial Intelligence is enormous. I chose to focus on 3 fields: medicine, work and music. Each time, I met the humans behind the algorithms.

0'49'' AI is already present in the machines that assist doctors, especially technical aspects such as medical imaging.  It also helps more and more  for the treatment of cancers by accurately analyzing tumors for example.

1'02'' Some start-ups are developing  artificial diseases that are very practical for patients. This is what the Lausanne residents of Biped do.

1'09'' This system helps visually impaired people, like Ali, to navigate. It is equipped with 3 cameras and a small onboard computer that allows the AI to operate in real time.  Each use allows the machine to  improve  and improve.

01'29'' Maël Fabien, Bipep CEO

T he cameras film the environment. O n is able to detect text, we are able to detect the number of a building too, eventually pedestrian crossings,  and to inform about the type of obstacles, which allows Ali for example to make different decisions if he hears a bus arriving or an electric scooter. It's not the same type of reaction we're going to have.

1'48'' Today, Maël accompanies Ali totest the latest updates and progress of his Artificial Intelligence.

1'55'' Ali is blind and immediately joined this system.

2'01'' Ali Yusuf

If you want the cane detects obstacles but at too low a level. And Biped, suddenly, comes to compensate for this lack.

2’10’’ Journalist

And what does it feel like to be led by AI, by artificial intelligence?

2'15'' Ali Yusuf

Frankly,I think it's scary at first, anyway.  You say, she's going to control me, she's going to hack me stuff like that, but in the end it's just frankly trusting the technology and especially the people who developed it and telling yourself that it's a safe device and go for it.

2'30'' It is in these offices of the Biopôle scientific campus in Lausanne that the system was designed.

2'37'' Maël Fabien, Bipep CEO

This is where everything is done. So all the software development is done here for artificial intelligence. We are 5 in total developed the software. We have all the storage with the components we use for assembly. We have an assembly line where we are  assembling a few units. That's more the meeting aspect and we even have the small servers  where we run the calculations and where we drag our algo in the back of the room. So yeah, we really do everything in this office.

3'05'' Mael's team has been working for 2 years on the development of artificial intelligence at the heart of their system. To train it you need data and there are several ways to get it.

3'18'' Maël Fabien, Bipep CEO

Data that we will be able to collect during the use of the device. And then after, there are also simulation platforms that exist where there are directly small environments like video games that will be quite realistic and we will be able to generate hundreds of thousands of quite complex cases.  Andif not, after  that,  there are some that have existed for a few years for the autonomous vehicle.  Mais then, it is licenses that usually cost, t it starts at half a million francs for just access to data, andthat is usually out of reach for our budgets.

3'47'' The start-up was lucky  enough to be approached by a major car manufacturer who provided them  with algorithms to improve their AI.

4'00'' It saved them time and money! The 2-year development of Biped cost 1 million francs,  and today, users like Ali, have to pay 129 francs per month to use it.

4’12’’ Quentin Bohlen, Journalist

Artificial intelligence is already omnipresent in the world of work: flow and machine management for industries, use of chat bots, even we Journalists, we are gradually replaced by text generators, ...

4'26'' Work-related burnouts  continue to increase, especially since the barrier between private and professional life has decreased with the advent of teleworking.

4'35'' A Swiss  start-up has developed anArtificial E Intelligence to give managers indications on the state of stress of its teams by analyzingthe voice during videoconferences.  It also allows each person to have feedback on their well-being at work.  I called Lara Gervaise, the engineer behind this system.

4’53’’ Lara Gervaise, Virtuosis Co-founder

Al ors, Virtuosis, is an application that analyzes the voice during conversations like ours to give feedback on well-being,stress levels and also to act as a communication coach.

5’12’’ Journalist

Now, Lara, I'm  asking you questions, will the AI think you're stressed or not?

5’18’’ Lara Gervaise, Virtuosis Co-founder

We'll see, I wonder,  I don'tfeel particularly stressed but maybe because I do.

5’31’’ Lara Gervaise, Virtuosis Co-founder

So here we can see my results of our call a little earlier. We see in particular that I was neither stressed nor anxious about our discussion and that I did not get angry, I did not show anger. It gives me anoverall score of 88 out of 100, so who is not, who is not too bad.

5'52’’ Journalist

And what can a team manager do, for example, with theseresults?

5’57’’ Lara Gervaise, Virtuosis Co-founder

So a team manager will not access individual data, only aggregated data  in  a dedicated App in which he will be able to see the overall level of well-being ofemployees. Also their risk of Burn Out.

6'15'' Lara is a Machine Learning Engineer. It was during her research at EPFL that she developed the AI behind Virtuosis. To do this, she recorded hundreds of conversations and then asked psychological specialists to analyze them.  This allowed him to develop training models.

6’31’’ Lara Gervaise, Virtuosis Co-founder

Ce that it does, it is immediately extract different different parameters. So the  pitch, the  tone of voice, the frequency and about 400 parameters like that. And from these different parameters it tries to recognize patterns to be able to predict the results.

6'53'' A large part of the engineer's job is therefore to  collect as much data as possible in order to perfect his AI.

7'04'' Then, everythinghappens here, at the EPFL datacenter.

7’11’’ Journalist :

So Lara, this type of  place is really quite crucial  for you, for the development of your AI?

7’15’’ Lara Gervaise, Virtuosis Co-founder

Yes absolutely, that's where everything happens, only where we can store data, butwhere computers can do calculations that require a lot of power to adjust the parameters of artificial intelligence models.

7'31'' Each time Virtuosis is switched on, the user has the choice to share his data to drive the machine. The voices are then anonymized before being transformed into digital data.  A way for the company  to reassure customers about the risks of privacy breaches.

7'48'' Finally, back on February 13th to this concert by David Guetta.  That evening, he is accompanied by the voice of rapper Eminem... entirely created thanks to the IA.

7'59'' Today,  more and more artists are getting into it, even here in Switzerland, and in musical fields that are not at all electronic.

8'10'' Quentin Darricau has developed an Artificial Intelligence capable of accompanying him when he improvises.  I meet him in the middleof a training session.

8’20’’ Quentin Darricau, musician

Thisseries of numbers corresponds to the little melody I just played, basically.  Andso that the machine can arrive associated with this melody,  or what would look like this melody, to a  given accompaniment, and well  all this is collected in a file withhundreds,  and hundreds of lines since each note, new note generates a line. Andso when the AI is going to know something that looks like a piece of this improvisation, it will trigger this chord, this accompaniment that I want, at the moment when I play this sequence of notes.

8'52'' The musician spent hours playing in front of his computer to get a result that satisfies him.

8'57'' are:

Hi, my name is Claire.

9'03'' He even went so far as to create a face to his Artificial Intelligence. Claire accompanies him during his public performances. If at first it was curiosity, little by little Quentin took a liking to playing with her.

9’18’’ Quentin Darricau, musician

Idid not necessarily expect it at first, and it is true that there are times when I let go completely and then I am completely caught up in the sound that comes out in the music and that's  it, it also exists, italso works with the machine.  

9'37'' Here, the musician is accompanied by thesacred polyphony of Claire, which would have been complicated to achieve with humans. Claire's creation allowed the musician to perceive artificial intelligence differently.

9’50’’ Quentin Darricau, musician

By really putting my feet in it, I have a lot of de-dramatizing what this object was, whichis neither more nor less than a tool that classifies data, ithas absolutely nothing frightening. It's still a tool like a hammer.  A hammer if you  plant nails with everything is fine if you hit someone down, it's  not okay but it's not the hammer's fault.

Total : 10’20’’