
Singularity: an artificial intelligence capable of dominating man could occur within six months
The possibility of an artificial intelligence dominating the human intellect – the famous technological singularity – is not unanimous. While many researchers project it in several decades, the CEO of Anthropic, it, it is possible within six to twelve months, in its own terms. With regard to his position in the world of AI, he is obviously not one of which we can ignore the words. Just like Sam Altman, The boss of Openai, who had expressed himself on the subject at the start of the year.
Predictions that have been accelerating since the arrival of major language models
It is in particular the major models of language (LLM) that have rebatted the cards. A recent large -scale analysis, relating to the predictions of 8,590 experts, shows a shift: IA researchers are now considering general artificial intelligence (IAG) around 2040, against 2060 BCE. More optimistic, entrepreneurs in the sector place for their part on 2030, still according to this analysis. It must be said that current systems, like GPT-4, impress by their ability to process complex requests or generate relevant responses. Their billions of learning parameters give them a large field of skills, from translation to content creation.
Technical acceleration engines
What feeds this potential acceleration? The document highlights Moore’s law: this principle wanting the double computing power every 18 months approximately would bring AI models closer to the capacities of the human brain, believe the analyzes. An AI dealing with information so quickly and although man could logically overcome it in various fields, from reasoning to the analysis of massive data. Another track, quantum computer. Still to its stammerings, she promises, according to specialists, calculations out of the reach of current machines. What to exponentially accelerate the training of neural networks and, perhaps, paver the way to singularity, always according to the optimists.
Sam Altman is never the last to speak of general artificial intelligence. © Shutterstock/Rokas Tenys
Persistent technical and philosophical obstacles
However, the study also returns to the numerous obstacles, both technical and philosophical, which still stand on the road which leads to AG. Indeed, current LLMs excel in simulating understanding of language. But human intelligence, with its emotional, intuitive or creative components, would remain out of reach, according to some of the respondents. And some voices count more than others. Thus, Yann Lecun, one of the fathers of deep learning, is skeptical: for him, AI will never be able to fully replicate human intelligence, too specific. He even proposes to speak of “advanced artificial intelligence” rather than IAG, self-awareness, for example, completely escaping current technologies, he points out. Moreover, the text recalls that certain models of the human mind count up to eight types of intelligence, such as interpersonal, logico-mathematical or existential intelligence. For skeptics, it is even possible that AI never manages to emulate all of these intelligences.
Issues that make you dizzy
Now let’s move on to the part that asks the most questions: faced with a potential superintendent, the ethical and societal implications are simply colossal, according to the document. How to control an AI superior to humans and capable of autonomy? It is an obviously fundamental question for researchers. The text reports a consensus among the experts cited: ethics must be in the foreground, with strong regulations to guarantee beneficial use of humanity, or simply its survival if things are going as badly as the fiction has been predicting it for decades.
It will also be for the company to anticipate major upheavals in employment, education or health. And if the singularity quickly happened, the projections indicate that it would take immediate adaptation plans to amortize the socio-economic shock. From our point of view, when we see the fears and the upheavals aroused by the current “simple”, it is difficult to imagine that a plan could be effective in the face of the chaos that an IAG would be if it had to disembark tomorrow.
Want to save even more? Discover Our promo codes Selected for you.