Artificial Intelligence is a fashionable topic all over the news. Have you taken the time to think deeply about it? Gaspard Koenig did not only think about it, he also traveled over several months to meet with 100+ thought leader in the field. He interviewed more or less everyone you could name on the topic through America, China, Israël and Europe.

He released in 2019 a smart essay under the name « La fin de l’individu ». It is a philosophical reflexion on Artificial Intelligence (AI) and journalist style account of his interviews. Coupled with a pinch of history, economy and sociology perspectives. I recommend the book if you want read something smart and deep on AI.

Résumé (en Français) or synthesis (in English)?

I like the summary on the http://www.decitre.fr website. And you can order the book on the link below. I have it on Amazon Kindle but preferred to advertise Decitre.

https://www.decitre.fr/livres/la-fin-de-l-individu-9791032907207.html

« Quel avenir pour l’individu et ses libertés à l’ère de l’intelligence artificielle ? Pour répondre à cette question urgente, Gaspard Koenig a entrepris un tour du monde de San Francisco à Pékin, d’Oxford à Tel Aviv et de Washington à Copenhague. Il a rencontré plus de 120 professeurs, entrepreneurs, intellectuels, politiques, économistes, artistes, et même un magicien. Au fil de ce périple émerge une véritable philosophie de l’intelligence artificielle (IA). 

Celle-ci ne menace pas l’existence d’Homo sapiens et les robots ne voleront pas nos emplois. En revanche, en déployant des techniques d’optimisation, de prédiction et de manipulation à grande échelle, l’IA remet en cause le fondement même de nos Lumières : l’idée d’un individu autonome et responsable. L’intelligence artificielle nous prépare ainsi des droits sans démocratie, un art sans artiste, une science sans causalité, une économie sans marché, une justice sans coupable, des amours sans séduction… 

à moins que nous ne reprenions le contrôle en forgeant pour nous-mêmes un droit à l’errance. Un récit philosophique pour notre époque, fourmillant d’informations, d’anecdotes, d’états d’âme et aussi d’humour. De quoi rendre l’IA plus facile à comprendre et, espérons-le, à maîtriser ! »

I happened to write a synthesis in English while as I was reading the book. I wrote it for my own use in order to remember some concepts rather than forget it all once I had moved to another work. The synthesis may not be easy to understand without having read the full book. And as every synthesis it is truncated and does not do justice to the author deep thinking. Also the book is in French and I am not sure it exists in English. I shared it below. Read it only if you do not plan to read the book!

Artificial intelligence is not a threat to mankind: AI needs us and lacks common sense

Prophets of doom predict a super artificial intelligence evolves and take over the world. Think about the Matrix movie for a simple example. After a range of interviews on this topic, the author concludes AI will transform our relation to the world but it will neither replace human beings nor take over them.

AI requires a lot of human based work to feed it. It requires coding if it is based on brutal force to calculate the best decision (Good old fashion AI, think the way IBM won against Kasparov at chess). It requires a lot of organized data if it is supervised learning (i.e. one feeds dozens of thousands of cancer tumor radiographies until AI spots it better than human beings), and a goal to be defined if it is based on reinforced learning (i.e. one defines what winning at the go game means and lets AI play millions of time until it gets really good).

And AI lacks common sense. It may play go better than us but needs to learn it again if it plays on a 18×18 goban instead of the traditional 19×19. Thus it is not really intelligent: it merely replicates the results of intelligence. AI lacks the basic human ability to adapt to new situations. And it is unable to define its own goals, it only learns what to do in order to reach them.

Artificial Intelligence is a threat to individual freedom and is politically oriented

Artificial Intelligence coupled with behavioral economics create new situations of permanent nudge. The the human brain is particularly permeable to nudge. We are also much less rational than we would like to admit. Professional nudgers are taking advantage of it. Thus, we may have the feeling to make free choices while we are actually funneled into predictable actions. This poses a significant threat to individual freedom. Both as a concept and as a daily life reality.

Artificial Intelligence is not politically neutral. Algorithms maximize an equation with underlying objectives. The objective may be to make us spend more time on a video platform, to have us spend more money, or to maximize common good. But there are many ways to define the common good. AI will optimize the common good as it was defined by the coder.

The definition of common good may even push the AI to channel an individual towards a suboptimal situation in order to maximize the larger common good. If this means Google Map delaying you by a few minutes to balance car density on a highway, it may not be such a big deal. If this is a government AI deciding about your university path, this may have higher impact. And there is no easy way to know what the AI in front of you is solving for.

AI is a new paradigm to live with, changing our relationship to the world

AI is a new paradigm we will have to live with. The author makes a parallel with the concept of equality of human beings that has risen over the last four hundred years and more or less imposed itself. Considering all human beings are equal had a deep impact on domains such as art, science, economy, justice, politics and religion. Similarly AI may have impact on these fields.

Art – AI art would mean algorithms to create music or painting in a specific style and would change the relationship we have with an artist work and evolution. An AI artist would not be young or old. It would not die and its work would never be complete.

Science ⁃ Machine learning based science allows machines to understand the laws of physics without equations or logic, just by observing what happens. This may bring back some magic in science as things happen because they happen and it may not be necessary to conceptualize physics as much as we do.

Economy ⁃ AI dispatch may create perfect markets without the need for a regulation based on price as the typical liberal market assumes. Thus, free market economies where demand and offer are free to fluctuate may not be the best path towards economic development anymore

Justice ⁃ Justice may have to take into account the algorithm responsibility and the nudging effect. We all know the dilemma of the responsibility if an autonomous car creates an accident. It is not obvious to define responsibility between the car owner, manufacturer, AI coder, etc. And it may also have to take into account when an individual can claim he or she was nudged into doing something

Politics ⁃ Autocracies may be better placed to feed data into AI algorithms and impose AI based societies, thus getting an edge over liberal democracies. It may prove wrong the assumption that economic welfare is linked with Occidental style democracies

Religion ⁃ AI could become the daily companion looking over our shoulder like God used to be in the Middle Age. As people share their emotions and post pictures on the cloud several times a day, their behavior is not very different from the daily devotion that could be observed centuries ago. AI could fill the individual void in each of us.

3 continental models in competition: China, Europe, US

3 continental models appear to compete on AI: Confucean China, Stoic Europe and Protestant North America. In China individual privacy is not an issue and AI is a way of life. China appears very well placed to make the best of AI and is described as counting on it to get it global supremacy back. Protestant America is described as torn between two worlds: it assumes it can make the best of technology while still retaining individual freedom and privacy. And Europe may be committing digital suicide by preferring the moral honor to protect individuals while crippling its AI development through GDPR laws.

Individual data ownership and Prime Directive as the way forward

Finally Gaspard Koening preaches for the right of individual to own their data and define their « Prime Directive » (indeed, a Star Trek reference). It may be the way to preserve individual freedom and get Europe back on the global AI track without losing its soul.

As the individual owns their data, he or she is able to define a price for it. This means we may have to pay for the digital services if we do not want to share our data. And we may get paid if we are a key content provider.

This Prime Directive would be the right to define the objectives AI should solve for you. For instance asking Google Maps to bring you through the fastest track, the most beautiful road or the path where you are most likely to make interesting encounters if you are on vacation. Digital giants (think GAFAs) would have to accept the individual Prime Directive rules when dealing with individuals rather than the opposite.

Each of us would need to think deeply how he or she wants to be nudged. Artificial Intelligence would influence us and maybe manipulate us. But we would decide how we want to be manipulated. At this price fee men could continue to be free men and not digital serves.

Et si vous vous intéressiez un peu plus à l’auteur ? Gaspard Koenig articule une pensée précise et fascinante sur de nombreux sujets. Et pour une version alternative de synthèse anglophone, Gaspard m’a gentiment indiqué l’existence de : https://www.economist.com/open-future/2020/01/10/a-cry-for-freedom-in-the-algorithmic-age