Two strong initiatives led by the French government have the full attention of innovation actors. Through France’s strategic plan for artificial intelligence, it is obvious that the President of the Republic wants to help France recover its strong position in the field of technology, one it has unfortunately lost in the past decades because of inadequate policies. It is also under these circumstances that the idea of a European agency for disruptive innovation, based on the model of the US agency DARPA, is increasingly being mentioned.
“How can France catch up its delay?”
Founded in 1958 in reaction to the Soviets’ launch of Sputnik, this military agency’s goal is to create applied innovation based on fundamental research. Over the course of 60 years, it has accumulated many successes, mainly due to alternative strategies in innovation. It can be credited with the invention of the Internet (Arpanet, which later became TCP/IP), the GPS, its decisive contribution to artificial intelligence and multiple other inventions. One of the agency’s strengths is the quality of its governance and its independance from military tutelage, which allowed it to be both daring and perceptive. But to sum up what characterized DARPA before anything else, it would be its capacity to promote disruptive innovation to the detriment of traditional R&D models, its readiness to organize contests open to all, in order to respond to particularly ambitious issues. Since then, the Internet revolution has accustomed us to the emergence of creations from the “multitude” – Wikipedia, Github, Tensor Flow, Linex to name but a few.
How can France catch up its delay? Emmanuel Macron’s statements, following the release of the report on artificial intelligence by MP Cédric Villani, demonstrate a willingness to implement public policies focusing on innovation, an effort which we can only applaud. There is no doubt the United States, likewise in China and Israel, has seen its ecosystem grow and benefit from the generally quiet support of the political sphere. If the intention is praiseworthy, we must pay careful attention to the method. Yet, most of the measures involve a framework very similar to traditional R&D. Many of his proposals are designed as if French research institutions were the perfect environment to foster the development of digital innovation. If the Villani report in itself flags insufficiencies and proposes to create an innovation framework similar to DARPA, the question remains of whether, 60 years after the Americans, any of this is still relevant. Indeed, if there is one lesson to be learnt from the history of innovation, it is that innovation rarely offers a second chance. Attempts to catch up usually turn out to be unsuccessful and recycling previous organizational models almost systematically leads to failure. At a time when the necessity of a European DARPA is mentioned in all places, why are we not trying to reflect upon what our next move ought to be?
“Making the choice of the famous serendipity”
In this respect, what can we observe? A world in which the structuration of value is rapidly evolving. The economist Philippe Aghion thus observed in his recent work that States are only partially capturing productivity gains resulting from their own innovation, especially because their researchers share their work for free. For example, Yann LeCun, one of the leading scientists in artificial intelligence, observes that he spends more time interacting with the “multitude” (thousands of innovators gathered on the TensorFlow platform dedicated to artificial intelligence) than performing research in the traditional sense. This dynamic seems to be lacking in France: in China, being involved on platforms like Github is almost the equivalent of taking part in university work. Taking part in an “innovation challenge” – such as the one proposed by Kaggle – is also considered to be a necessary step for those wishing to pursue serious research in AI.
This is what’s missing in the current dynamic, be it regarding artificial intelligence or innovation in general.
The fact that Stack Over flow, Github, TensorFlow or Kaggle aren’t mentioned as collaborative platforms in the Villani report reflects a vision that may be out of date. Yet, while it becomes obvious that digital technologies carry the risk of creating a schism between people who are highly educated and those who are not, between the stock options beneficiaries and the gig economy subcontractors (an economy based on new forms of employment fostered by digital platforms), we should actually place the opposite bet: that of massifying the distribution of knowledge and making the choice of the famous “serendipity”.
“This is about accepting the idea that collaboration between diverse communities can trigger extraordinary results”
This is no hypothesis based on altruism, but a pragmatic necessity. It is now clear that innovation is increasingly sparked by innovators coming from environments outside of academia. Platforms should therefore be used as the next tool to foster innovation, especially at a time when they have access to unrivaled group facilitation technologies. Beyond that, this is also about betting on the democratization of innovation as a way to overcome the obvious pitfall of digital technologies. Indeed, there is a strong risk that the latter benefit only an elite, who, despite all the rhetoric deployed on the issue fails to change the world in a broadly inclusive way. This is not to deny the importance of researchers: they do have a crucial role as crystallizers and facilitators of innovative groups. This is rather about accepting the idea that collaboration between diverse communities can trigger extraordinary results, and that this is exactly what the aforementioned collaborative platforms allow for.
As a result, one should be particularly careful when it comes to public expenditure: €1.5 billion is a significant amount (although it remains minor compared to private investments in China and elsewhere). Thus, the creation of “AI coffees”, ephemeral places dedicated to the demystification of these technologies on a large territorial scale, would only cost a fraction of the amount considered in the AI plan. Similarly, creating awards for popular actions around artificial intelligence would certainly be both virtuous and efficient. Indeed, there is a possibility that these actions could have results incomparably superior to those of conventional initiatives. For it becomes difficult to deny that taking alternative paths, while it might be riskier, has more chances of sparking unexpected outcomes.