You left for California in the 1980s…
I was in the United States at the beginning of the 1980s – I was working on my thesis in Los Angeles – and it’s only when I returned to France that I realised the gap between France and the U.S. in terms of computer science at that time. To give you an idea, when I lived in the U.S., I often sent emails to other researchers all around the world. At the same time, in France, the few researchers who used emails most often could only use it within their organizations. As for the Web, it did not exist yet.
How did you get to Stanford in 1995?
I came back to the U.S. for personal reasons: in fact, I followed my wife, who was going to develop a branch of a French start-up in Palo Alto. I quickly found a job as guest professor in the computer science department at Stanford.
It was during the dot-com bubble, wasn’t it?
Stanford was filled with a wonderful dynamic. In my research group, several start-ups were created every year – for example, one of the postdoctoral researchers of the time, Anand Rajaraman, created a price comparison tool that was quickly bought by Amazon. Being at the core of this ecosystem, I was certain this was Stanford normality. My older colleagues explained later on that it was actually a golden age… I was incredibly lucky!
In 1996, you were there during the first demonstration of the search engine that would become Google. A demonstration led by Larry Page and Sergey Brin (then postdocs in your research group). Was it impressive?
I need to say that we were expecting this news because Brin and Page had already explained the concept in small-committee working groups. But to be honest, I still was blown away by the quality of their search engine results. At the time, we were used to the mediocrity such results, but with Google, it really was different.
About Brin and Page, I think it is important to debunk the myth of two self-taught men in their garage who found a revolutionary algorithm on their own. They were not self-taught, nor were they in a garage! They were two brilliant students working in Stanford premises, and got help from the university and from professors who would lend them the rooms and machines they needed. Also at the time, industrialists came to our working groups all the time (when in France, I rarely saw any!), it was an enriching exchange for everyone. It was the heart of an effervescent and highly stimulating ecosystem.
In fact, Brin and Page found their original idea in a paper by John Kleinberg from IBM. Their brilliant idea was to take Kleinberg “HITS” algorithm, to simplify it, and to apply it to web scale so it could sort millions of pages (we weren’t talking about billions yet). We must acknowledge that Brin and Page, besides modifying the algorithm, had great ideas, such as using clusters of basic computers rather than to rely on powerful servers, and to rely on the cheapness of memory.
Did you foresee Google’s success at the time?
Could we tell if it was going to work? Yes, we already knew it. Could we foresee such a success? No, because the success did not come only from the search engine… Very quickly, they built their business model on advertisement, and this is where the money came from.
I told you we had not foreseen such a success, and it is true… However, there was a time when all the students and professors of our group wanted to go work at Brin and Page’s. Our group manager had to call Brin and Page to ask them to stop poaching us in order to ensure the group’s survival. Without this, many of us would be multimillionaire by now!
And on the contrary, looking back, does the scale of Google’s success surprise you?
They managed to do two very different things: they created a start-up and they built an empire – with all the new services they progressively added. As for me, I will tell you about what I know a little about, that is creating a start-up. They started with a blatant need: the Web was awesome but it was lacking something, a good search engine. They felt the urging need to find the right information in a haystack of information, and they developed a great technology. Personally, I think that Kleinberg’s algorithm, modified by Brin and Page, really is one of the most beautiful algorithms I know.
Knowing Brin and Page, do you think their motto “Don’t be evil” for their search engine was sincere when they created it?
To be honest, I knew Brin very well and Page very little. Brin was a very creative man: he barged into my office every week with new crazy yet brilliant ideas. He also was a true idealist. You need to know that “Don’t be evil” is not the official motto. It is “Organize the world’s information and make it accessible and useful.” Brin was someone who wanted to do good things, he may have been naïve, but to me he was sincere. Let us not forget that these two young men were 25-year-old students who could not afford a car, and in no time they had to manage a colossal fortune.
Personally, I thought for a very long time that their search engine bore no ill will. They started from a sincere vision but the investors turned Google into a company like any other and the search engine became biased.
It still makes me smile when I read this excerpt from their research paper announcing Google’s launch: “We believe the issue of advertising causes enough mixed incentives that it is crucial to have a competitive search engine that is transparent and in the academic realm.”
By the way… what is an algorithm?
An algorithm is a sequence of steps to perform in order to solve a problem. Whether you want to sort webpages or get dressed in the morning (avoiding to put on socks over shoes, for example), there are always several ways to do it, and each one is an algorithm.
Why do we talk so much about algorithms today when they have been around for thousands of years?
What changed is simply the fact that, since the last century, we have built computers able to execute them for us. And if, as early as the 18th and 19th centuries, scientists were already designing calculation machines, they had not imagined the network that would allow the collaboration of billions of computers.
Thanks to computer science, we found the way to have others do our work and that computer / algorithm association opens up many possibilities as to what we can create every day. One of the examples that struck me recently was the number of Whatsapp employees (about thirty) compared to the number of Whatsapp users (1 billion). This would not be possible without algorithms.
Why algorithms are scaring us?
It is mostly due to our culture. People in France are more sensitive to these topics than in Asia or in the United States, and the French media does not help because it highlights the issues algorithms raise. French people are also quite complex: they both want to protect their personal data and to keep downloading apps that communicate this very data to anyone. They like IT but bear the gnawing fear of what it could do.
Do we have to see algorithms as something to replace or complement humans?
The main fear is to lose a job because of its replacement by machines. For example, Amazon is looking to increasingly automate supermarkets, thus leading to jobless cashiers. But the real question is to know what people truly regret: their jobs or their revenues? If you ask me, it’s the revenue. If we replace all these people’s jobs by machines, the revenues of these machines must not end up on the account of bankers or such: it has to be redistributed. The question is then not technical, but rather political.
The second issue is a philosophical one. Indeed, until today, we associated our social life with working, which is considered a virtue. And we are talking about no work… because we will be replaced by machines. The current transformation seems brutal, but let us not exaggerate. We will have time to adjust because this transformation is going to take time: we are not talking 5 or 10 years, but much more until work disappears almost completely.
So, the best way to talk about this transformation is to compare it to the industrial revolution?
It is more a thought revolution that it is an industrial revolution. It is not only about accessing a new tool, because meanwhile we have to acquire a new way of thinking, we have to transform our minds to use it. I think we should compare the emergence of algorithms to the emergence of writing rather than that of steam.
Hierarchical organizations, whether companies or governments, are not up to date for this tool, and will have to adjust. This major change affects many fields. For one, the way we do science has evolved a lot: for example, data analysis has greatly modified what climatologists do, and in biology, DNA can be studied like a program.
Artificial intelligence is getting a bigger place in the public debate. Is it a prolongation of algorithms or something entirely different?
For me, AI is not a field per se, it is a selling point rather that science. Since its beginnings, computer science has had a single goal: to solve problems with machines, and as a consequence to perform tasks in the place of humans. These machines use what we incorrectly call intelligence. In machine learning, thanks to a learning phase, we can take the same algorithm and, according to the data we feed it, it will change. I think it is awesome to solve problems we could not solve before, such as visual recognition. The game of Go is another great and recent example of this: we used a learning algorithm that analyzed the games of several players before learning to play against itself. In the end, it beat the greatest champions. It is not a new science, it is still computer science. We do not create another science every time we discover a new type of algorithm.
Is a technology inherently neutral? Or, in other words, what are algorithm developers responsible for?
Of course, it is the way we use the technology that is evil, good or neutral. I dismiss both extreme points of view on this topic: the “unicorn” version saying “we will deal with all of this,” and the “Frankenstein” one that we can understand as “Google is spying on us, Facebook is turning our kids into zombies…” The truth is, algorithms are neither good nor evil. They are written by humans and they can help us live better or not. We are now beyond the point when the responsibility is no longer the privilege of computer scientists: the whole society is responsible.
To illustrate this, I want to go back to the Google case we discussed earlier, which is now a problem in itself because it can create huge competitive imbalances. The blame is not really on Google engineers, but the responsibility is now for society to set rules, and for citizens to use or not this search engine. The only thing we can blame Google for is that they pretended their search engine results were neutral. Personally, I now use Qwant, which reminds me of the early-2000s Google… The choice is ours!
Willing to learn more about Artificial Intelligence ?Discover the first season of our story about AI !