LNovember 27 marked the second anniversary of ChatGPT, with an announcement that will not leave you indifferent: the launch of its new artificial intelligence (AI) model, Orion. It is hailed by its designers as a major advance towards artificial general intelligence (AGI, capable of performing virtually all human tasks). But, while its potential fascinates, its dangers divide and raise questions about the future of humanity.
Sam Altman, CEO of OpenAI, which offers ChatGPT, has never hidden his deep ambitions. According to him, “Artificial general intelligence will be the most powerful tool ever created by humanity, capable of solving problems that we do not even dare to consider yet”. Orion is a step closer to this vision, positioning itself as a model that aspires to a more holistic understanding of the world, unparalleled adaptability and an ability to empower oneself in learning.
The perspectives opened up by Orion are immense. With its deep learning capability 100 times more powerful than GPT-4, it revolutionizes the way we solve complex problems, anticipate needs and push the boundaries of innovation, redefining new horizons for artificial intelligence.
A frantic race
However, behind the enthusiasm lies a growing shadow. The cycle of transformation of artificial intelligence is accelerating: where it took several decades to observe a significant leap, these developments can now be counted in years, even months. The breakneck pace of these advances is raising concerns among experts.
Read the survey | Article reserved for our subscribers Artificial intelligences, the thousand and one ways to derail them
Read later
David Kokotajlo, a former OpenAI employee responsible for anticipating advances in AI, recently sounded the alarm. He who estimated in 2020 that the IAG would not see the light of day before 2050 is now revising his forecasts. According to him, there is now a 50% chance that an AGI will be available by 2027, only three years from now. But his predictions don’t stop there: he also estimates the probability that this advance will lead to catastrophic consequences for humanity is 70%. With Orion, the question of control becomes crucial. This ultra-sophisticated model could, by becoming more autonomous, exceed the initial intentions of its creators.
Read also | Article reserved for our subscribers Is ChatGPT a better parent than us?
Read later
Critics fear that the quest for efficiency at all costs opens the door to irrevocable effects: manipulation of information, disappearance of millions of jobs, even dystopian scenarios where humanity loses control in the face of an AI that has become uncontrollable. The most pessimistic point out the lack of safeguards in this frantic race. The shorter these cycles become, the more the risk of a drift increases.
You have 18.65% of this article left to read. The rest is reserved for subscribers.
