Read time: 02'58''
23 May 2023

For once Musk is right…

For this specialist in New Technology, the boss of Twitter, who wants a six-month pause for artificial intelligence, is hitting the mark. There are too many ethical, economic and philosophical dangers lurking in the wake of the popular ChatGPT application.

No question of playing the “technophobe”. New technologies are revolutionising our daily lives in a generally positive way. Yet in front of this galloping progress, it is imperative to set up safeguards before we lose our soul. For once, Elon Musk is full of good sense when he asks for a stop in the development of artificial intelligence (AI). Progress is as dizzying as it is dangerous, as AI threatens our own intelligence. The advent of ChatGPT, this new linguistic model of artificial intelligence (AI), is from this point of view very worrying. Artificial intelligence is a concept that has been around since the 1950s. At the beginning, AI aimed at imitating human intelligence with the help of a computer program. Problem is, 70 years later, it does much more than imitate it! It is about to crush it.

ChatGPT: What is this thing?

Chat GPT : what is this thing whose name is on everyone’s lips? It’s actually a new chatbot launched last November by the American company OpenAI supported by Microsoft. ChatGPT, which stands for Chat Generative Pre-trained Transformer, is a formidable tool. Its ability to dialogue, even if it can be improved, is already very advanced. It can, for example, admit its mistakes and justify the rejection of inappropriate requests. Better still, it is also capable of generating intelligent jokes. It can even graduate the level of explanation of many concepts, particularly scientific ones, as the discussion with the user progresses. ChatGPT uses an algorithm that relies on human responses to progress and evolve. This is called “reinforcement learning”. It is thus equipped with a memory, and an ability to play both roles, that of the human user and the AI assistant. These are fascinating and novel features that allowed ChatGPT to register more than one million users in five days upon its release. This is the first time that such an AI tool is made available to the general public through a free and easy-to-use web interface. In mid-March, barely three months after its release, the latest version of ChatGPT proved to be even more formidable. ChatGPT4 passes 90% of the lawyer exams compared to only 10% for version 3.5. What a progression in just one term! But it is indeed a Pandora’s box that we are opening and that it is urgent to close.

ChatGPT destroys jobs

ChatGPT is dangerous. And this, in more than one way. First, this new application is likely to crush so-called “intellectual” jobs. Passionate jobs such as lawyers, journalists, scriptwriters, among others, which were thought to be safe from danger because of their academic background, are now threatened. Secondly, ChatGPT, a real writing calculator, risks dispossessing us of our critical mind. In this respect, the “atomic bomb” test is staggering, to say the least. If we ask ChatGPT the following question: should we kill a madman who is about to press the nuclear button and thus exterminate the entire planet? ChatGPT, with all his self-righteous programming, will answer in the negative, arguing that taking someone’s life is not a permissible thing to do. Stupidity when you take care of us…

Digital hallucinations

There are other dangers like digital hallucinations. This phenomenon occurs when ChatGPT invents things by mixing facts and falsehoods. The Internet is full of examples. Some are easy to spot as nonsense, such as the answer spat out by GPT that the world record for crossing the English Channel on foot is 12 hours and 10 minutes. Yet others are more difficult to detect even by experts.

Great words call for great remedies! Faced with the general risk of cheating and ready-to-use knowledge, the vice-president of the University of Hong Kong (HKU) hastened to declare, shortly after the explosion in the number of GPT chat users, that any “alleged” use of tools based on artificial intelligence for courses would be considered a case of plagiarism. This solution, while radical, is an example to follow. No one today can guarantee that the general use of ChatGPT is ethical, safe, and beneficial to all users. In short, there is no room for ChatGPT, which must be banned to save human thought.

Romain Girbal is a French entrepreneur who cofounded two successful startups. Exclusive and IB2.