Yesterday, I tried asking the new version of the ChatGPT generative artificial intelligence software (number 4) what it thought the top three problems of the world in 2050 would be, and the top three advances. In a few seconds, I saw a very articulate text being written, starting by apologising for not being able to predict the future; then explaining that, taking into account the latest discoveries, the three main problems in thirty years’ time will be climate change, epidemic risks and cybercrime. And that the three main advances will be in the conquest of space, in health, and in the control of greenhouse gas emissions.
Anyone can easily engage in such an experiment; already, many students have understood this and use ChatGPT to prepare lectures, essays, to draw up thesis plans, or to write texts in the style of Stendhal or James Joyce. They can pretend to be the author, or to have found a missing work by an author whom the software has imitated. For the moment, these artificial intelligences are not yet creative, nor are they really useful. Their progress is fascinating, however, because they are constantly learning from everything they do and how they are used.
Very soon, future versions of this software will be able to replace humans in a huge number of tasks. The jobs most at risk will be highly skilled ones such as lawyers, translators, accountants, financial analysts and journalists. According to recent studies, a quarter of all jobs in the United States and Europe could very soon be replaced by generative artificial intelligence systems. And then there will be consultants of all kinds, engineers, architects, video game designers, software creators. Once again, technical progress will proletarianise activities requiring very high qualifications.
All of this is nothing compared to what will await us a little later: each of us will one day be able to create at will, thanks to a later version, (number 25 perhaps) children’s stories, novels, philosophical essays, musical works, sculptures, photographs, films, software. Even more than that, it will be possible to entrust this software with the task of inventing other software that will itself be able to create other increasingly creative software.
Everyone will also be able to delegate these creative activities to an artificial intelligence that will be their intellectual and artistic assistant. This will be increasingly sophisticated, and here again, money will allow some people to have increasingly sophisticated assistants of this kind.
This software will then be able to take initiatives. To begin with, they will be able to write emails in your style, using your mailbox, or that of a private or public leader, to give orders, reveal secrets, order embargoes. By multiplying, they will be able to create indescribable disorder. And worse still, artificial intelligences could join forces to use such applications against humans, or against humanity in general.
All these dangers are real. And much more imminent than we think. There is no way to avoid them, or to ignore them. Nor should we be content to marvel at the innocent games that these software programs now allow, nor at the formidable means they will put within our reach to better reflect, create, progress and enchant.
A truly global charter should be put in place as soon as possible, recognising the hopes that these artificial intelligences hold for us, but prohibiting them from harming their human authors. This is what the great American science fiction writer Isaac Asimov proposed in 1942. Naturally, we will not do this.
Just as we have not done so effectively for taxation, nuclear weapons or genetic engineering.
Unless we finally recognise, for this subject as for so many others, the urgent need for a planetary rule of law.
Painting: René Magritte, The Empire of Lights, 1954