Automated Inequality
Harvard IOP

Automated Inequality


International Organizations

Institute of Electrical and Electronics Engineers (IEEE)

The AII, via Cyrus Hodes, is participating in the IEEE Global Initiative for Ethical Considerations in Artificial Intelligence and Autonomous Systems. In particular as a member of the General Principles and Guidance Committee. This Committee is working on the overarching principles of the ethical design and use of autonomous and intelligent systems, focusing on the broader issues related to the acceptance of these systems by the general public.

August 28, 2016, The Hague, Netherlands: Symposium on Ethics of Autonomous Systems (SEAS Europe)


We are the useful idiots of Artificial Intelligence

Artificial Intelligence (AI) is the main economic and political issue of the 21st century. Its regulation and governance will greatly impact our future. After unrealistic promises in the 1960s and 70s, disillusionment was immense since computers of the 80’s remained primitive. We had to wait until 2013 to witness the revival of AI ​​with the implementation of software surpassing humans in the visual recognition field.This AI awakening ​​is linked to the deployment of Deep Learning by Yann LeCun at Facebook. Deep Learning is a technique that allows a program to learn to represent the world. It learns through a network-based architecture of "virtual neurons" that perform elementary calculations, bringing it a little closer to our own brain functioning. However, unlike a child, AI needs many examples to learn. The key success factor is no longer the length of computer code, but the size of the databases. As of now, AI ​​needs to see millions of pictures of animals or cars to recognize them. GAFAMI (Google, Apple, Facebook, Amazon, Microsoft and IBM) and BATX (Baidu, Alibaba, Tencent and Xiaomi), which are Silicon Valley’s Chinese competitors, enjoy an overwhelming superiority with installed bases of billions of customers who feed the databases. Consumers are the useful idiots of AI: our lives are devoted to educating AI even if we are not aware of it. We publish nearly ten billion images every day on Facebook that continuously improve the visual recognition of AI. Optimization of AI ​​also requires special microprocessors. The learning phase of the neural networks preferably uses so-called "GPU" processors initially designed to operate the video games. For the response phase, digital giants develop dedicated processors that they have a monopoly on. Google uses the Tensor Processing Unit (TPU), which won the game of Go against the human world champion, and Microsoft uses the Field Programmable Gate Array (FPGA). We are heading towards an oligopoly of a dozen US and Chinese companies controlling AI ​​thanks to both giant databases and proprietary microprocessors not accessible to other companies. The AI ​​giants are building ecosystems around an "AI tap" that they alone possess. Other companies will be able to plug into AI ​​modules to process their data, as they can now place their programs on smartphones via Apple and Google-Android app stores, but they will not have full control. Thus, it is possible to benefit from the AI image ​​"Made in Google" through their "Tensor Flow" architecture provided that one accepts being a digital vassal. Andrew Ng, the head of AI ​​at Baidu (the Chinese-version of Google), explains that AI will become as crucial as electricity. It will probably be necessary to consider the creation of a public utility of AI ​​with a ban on cutting off its supply, the same way it is now forbidden to cut electricity to a company or a household.

By Laurent Alexandre, Senior Advisor at the AI Initiative