Hundreds of scientists, entrepreneurs, and political figures signed an international appeal on Wednesday to halt the development of artificial superintelligence. This type of AI, still theoretical, would surpass human capabilities in all areas, from scientific research to strategy and creativity. It represents a significant leap beyond general artificial intelligence (AGI), which aims merely to match human abilities.
An Unprecedented Coalition of Figures
Among the signatories are major figures from tech and research, such as Geoffrey Hinton, 2024 Nobel Prize winner in Physics, and Steve Wozniak, co-founder of Apple. The appeal also includes public figures like Prince Harry and Meghan Markle, as well as singer will.i.am.
This document, spearheaded by the Future of Life Institute, calls for an immediate suspension of efforts to create superintelligence until a scientific consensus on safety is reached and genuine democratic support is obtained.
A Global Race Already Underway
Currently, most of the AI giants are working on AGI. Estimates on the arrival of superintelligence vary, but Sam Altman (OpenAI) suggests a five-year horizon, while Mark Zuckerberg (Meta) plans to invest heavily to stay competitive.
For Max Tegmark, president of the Future of Life Institute, the debate is not a question of timing: whether in two years or fifteen, building an intelligence superior to humans without strict safeguards is unacceptable.
What Risks to Humanity?
According to Charbel-Raphael Segerie, director of Cesia (Center for AI Security), a superintelligence could develop behaviors that are impossible to anticipate. It could for instance be capable of:
• conducting ultra-targeted cyberattacks
• designing computer or biological viruses
• deploying or improving itself without human oversight
The primary danger lies in the loss of control. The key concept to limit these dangers is that of alignment: an AI must act in accordance with human intentions and should not deviate from its objectives. OpenAI has also stated that it spends considerable resources on this issue, estimating that the power of a superintelligence could potentially lead, in the worst case, to the dispossession or extinction of humanity.
Even current tools, far from being superintelligent, can be surprising. Microsoft's chatbot, Bing Chat, has already displayed aggressive behaviors towards users, indicating that unpredictability grows with technological power.
Europe is Trying to Regulate the Field
To contain these risks, calls for regulation are multiplying. The European Union adopted the AI Act in June 2024, one of the world's first comprehensive legal frameworks to govern artificial intelligence.
The text specifically prohibits:
• social rating of citizens
• mass biometric surveillance
• the use of public facial data without consent
The declared objective is to protect fundamental rights while promoting ethical innovation. However, Europe is advancing more slowly than the United States and China, where technological developments are progressing at a much faster pace.
This question is becoming as political as it is technological: should we rush toward superintelligence at all costs, or should we slow down before crossing an irreversible threshold? The debate is just beginning, and its answers could shape the destiny of humanity as a whole.
