(Image credit: Yuichiro Chino/Getty Images)
The emergence of artificial general intelligence (AGI) – an AI system with greater intelligence than humans and capable of solving a variety of problems efficiently – is a matter of time, not probability, according to a new study that includes the opinions of thousands of experts.
An updated analysis released February 18 by Cem Dilmegani, lead analyst at AIMultiple Research, looked at nearly 8,600 predictions from scientists, AI experts, and entrepreneurs between 2009 and 2023 to see when experts think this might happen.
The analysis section included 10 surveys involving 5,288 AI researchers and experts. The results showed that, based on the average of the data, there is a 50% chance that machines will achieve human-level intelligence between 2040 and 2061.
More recent surveys suggest that the technological singularity could arrive sooner. For example, one recent 2023 study of 2,778 scientists suggested that AGI will be reached no later than 2040. Some experts, like Dario Amodei, an AI researcher and CEO of Anthropic, suggest it could happen as early as 2026.
Progress in creating AGI has been driven by the rapid development of large language models (LLMs) based on transformers. This technology is used in chatbots like ChatGPT and image generators like Dall-E. Before these technologies emerged, some scientists predicted in 2019 that AGI could emerge by 2060, or perhaps never.
Why AGI Is Considered a Matter of When, Not an 'If' Matter
The analysis identified several factors that make scientists confident in the inevitability of the emergence of AGI.
First, unlike human intelligence, there is no theoretical limit to the increase in computing power. This is stated in accordance with Moore’s Law, which predicts a doubling of power approximately every 18 months. Future AI systems may one day reach human intelligence in terms of calculations per second – but only if this prediction is fulfilled. In recent years, many have expressed the opinion that Moore’s Law no longer holds.
Quantum computing is also mentioned in the study as a way to overcome computational limitations. Quantum computers are able to process calculations in parallel using the laws of quantum mechanics, while classical computers — including the fastest supercomputers — perform calculations serially. Thus, quantum computing could provide an advanced AI system with significantly more computing power than the best models available today.
However, some scientists in the field believe that further breakthroughs are needed to achieve AGI.
For example, Yann LeCun, Facebook’s chief AI scientist, argued in an October 2024 talk that Transformer-based architecture and current approaches to AI are incompatible with human-level intelligence. He also called on scientists to abandon the notion of AGI entirely. He believes there is a false ambiguity between its widely accepted definition and what an individual human can actually achieve — which in practice is a narrow set of specialized tasks, rather than the ability to learn any task.
Keumars Afifi-SabetSocial Links NavigationChannel Editor, Technologies
Kumars is Live Science's technology editor. He has written for a variety of publications, including ITPro, The Week Digital, ComputerActive, The Independent, The Observer, Metro and TechRadar Pro. He has over five years of experience in technology journalism, having previously served as
Sourse: www.livescience.com
