Six Months Later, France has Formulated their Deep Learning Strategy
Six months ago, I wrote that “The West is Unaware of the Deep Learning Sputnik moment”. It turns out mathematician Cedric Villani began a 6 month journey to learn all that needs to be learned about Deep Learning. He is the author of a report that France will use to drive their future Deep Learning strategy. Villani’s report is titled “For a Meaningful Artificial Intelligence: Towards a French and European Strategy”. Although Villani uses the term “Artificial Intelligence” to appeal to a much wider audience, he is actually responding to the “particular” developments of Deep Learning.
Cedric Villani in his speech at yesterday’s unveiling of the French strategy about the unexpected success of Deep Learning: “the world of neurons is not gone forever as previously thought. In an area where everyone thought was impossible, an algorithm that could play Go …The revolution is here, happening before our eyes.”
In his report, he writes about Deep Learning in 10 different occasions:
One of the reasons for the recent take-off of deep learning was the fact of GPUs (Graphical Processing Units) coming into general use… a GPU contains thousands of them, so dramatically accelerating the speed of calculations made and consequently of learning and data processing by machine learning algorithms, deep learning algorithms in particular. — page 51
One of the specific aspects of AI is the creation of de facto standards, notably of a technological nature: this is the case for deep learning for example, where technology such as TensorFlow (developed by Google) was adopted by an overwhelming market majority as soon as it was released. — page 59
The scientifically and economically dominant themes of the moment (learning, and the different strands of data science, or data analytics, also known as Big Data) need to be delved into further; semi or unsupervised learning, reinforcement learning, representation learning and domain transfer as well as unstructured data learning (textual data, tweets, blogs and other electronic media for example) also need to be on the programme, widening the scope beyond deep learning alone. — page 64
More broadly speaking, theory (and that of deep learning in particular) is lagging behind practice today, and close collaborations with other fields of mathematics and the information & communication sciences and technologies (without really being able to talk about interdisciplinary, then) are to be set up, from game theory to logic and formal proof, from information theory to geometric approaches. — page 65
Along with image and video analysis and vision processing, natural language processing has most likely gained the most from the arrival of deep learning. — page 65
Nowadays, our ignorance is principally due to changes in the paradigm that is introduced by machine learning, in particular deep learning. — page 114
But although certain models of machine learning are more easily explainable than others (systems based on rules, simple decision trees and Bayesian networks), nowadays their performance does not generally match up to that of deep learning algorithms. — page 115
The use of deep learning algorithms, which feed off data for the purposes of personalization and assistance with decision-making, has given rise to the fear that social inequalities are being embedded in decision algorithms. — page 116
In the first place, deep learning techniques are still too obscure (see above) and their audit protocols are still in their infancy. — page 117
Especially since, within the context of deep learning, data is used on a massive scale to produce correlations which could affect whole groups of individuals.
I repeat these parts here to give one an immediate sense of the impact of Deep Learning to the French AI strategy.
Cedric Villani is a brilliant mathematician and a member of French parliament. His research is also coincidentally extremely relevant to Deep Learning research. It is only recently that the theories of Optimal Transport theory has influenced work in Deep Learning (see Wasserstein metric and GANs). His other work in Riemann geometry and the Boltzmann equation may also be extremely relevant towards better theory in Deep Learning. Villani perhaps never trained a neural network, but he knows a lot more about Deep Learning than even an expert practitioner. In fact, I conjecture that the mathematics developed in the field of optimal transport theory may be the most mature mathematical approach to explore Deep Learning.
Here is a lecture by Villani worth watching:
The report contains the usual recommendations to develop competitiveness in any new technology. The sections however that are unique to the exponential growth of Deep Learning are the parts about “the Future of Work”, “Ethics of AI” and “Inclusiveness of AI”. Here are some additional interest quotes from the report:
About AI and humanity:
“Rather than undermining our individual paths in life and our welfare systems, AI’s first priority should be to help promote our fundamental human rights, enhance social relations and reinforce solidarity. Diversity should also figure within these priorities.”
About transparency and audibility:
“Transparency is clearly key, but looking beyond this issue, it is also vital to facilitate audits of AI systems. This could involve the creation of a group of certified public experts who can conduct audits of algorithms and databases and carry out testing using any methods required.”
“Artificial intelligence must not become a new way of excluding parts of the population. Public authorities could embark on specific programmes to support AI innovation in the social arena and provide the necessary systems for the various parties in the sector so that they can benefit from AI-related progress.”
I commend the French for placing humanity at the center of their AI strategy. They perhaps are behind with respect to technology, but they are way ahead with respect to setting the correct priorities.
The French strategy at its core boils down to these words spoken by President Emmanuel Macron in a Wired interview:
And Europe has not exactly the same collective preferences as US or China. If we want to defend our way to deal with privacy, our collective preference for individual freedom versus technological progress, integrity of human beings and human DNA, if you want to manage your own choice of society, your choice of civilization, you have to be able to be an acting part of this AI revolution .
This sentiment of the French is very plain and simple. Their conclusion is that who controls AI also controls civilization. This conclusion should not come as a surprise to anyone. It should be obvious because intelligence begat civilization.
BTW, I am still waiting for the US to start panicking. I’m not too optimistic, I don’t think that there are many “Cedric Villanis” (i.e. scientists) in the legislature or the executive branch of government. There is in fact an increasing probability that the US will shoot itself on its own foot with regards to an AI strategy. The recent debacles at Facebook, Uber, Tesla and the animosity against Amazon are newest signs that a backlash is brewing. The most productive thing we can do today is to emphasize the human benefits of AI and not simply drool over the latest developments.