AI Experts Warn of Possible Human Extinction as Risks Increase
There is growing concern among artificial intelligence (AI) experts that the development of superhuman AI could lead to the extinction of humans. In a recent study conducted with 2,700 AI researchers, it was revealed that 58% of them believed there was a 5% risk of AI causing dire consequences for humanity, including extinction. This largest-ever poll of AI researchers conducted at prestigious conferences sheds light on the potential dangers associated with the advancement of AI technology.
The report, authored by Katja Grace of the Machine Intelligence Research Institute in California, highlights the significant indication that the majority of AI researchers do not deem it highly impossible for powerful AI to wipe out humans. According to Grace, the perception of a non-negligible risk is more informative than precise percentage calculations. These findings raise concerns about the future implications of AI development and call for careful consideration of the risks involved.
However, not all experts are in agreement about the level of concern. Émile Torres of Case Western Reserve University in Ohio argues that many AI experts have a poor track record when it comes to predicting advancements in the field. While Torres acknowledges the lack of future-predicting expertise among AI researchers, it is worth noting that a previous iteration of the poll in 2016 successfully forecasted major advancements in AI. This divergence in opinion emphasizes the uncertainty surrounding the potential impacts of AI on our society.
Despite the differing viewpoints, it is evident that AI is progressing at a rapid pace. The survey revealed that AI systems could achieve significant milestones sooner than expected, with Silicon Valley’s push for widespread implementation of AI chatbot services serving as evidence. As early as 2022, AI is predicted to be capable of completing various tasks with a fifty percent or greater chance of success. These tasks range from producing songs similar to Taylor Swift hits to developing a payment processing website from scratch. However, it is important to note that certain operations, such as physically installing electrical wiring or solving complex mathematical problems, may take longer.
The study also unveiled forecasts for the future, with a 50% chance assigned to the development of AI surpassing human capabilities on any task by 2047. Additionally, there is a 50% chance that all human employment could be fully automatable by 2116. Compared to the previous year’s estimates, these projections are 13 and 48 years earlier, respectively. These timelines represent potential advancements that could shape the future of AI, but they are not set in stone.
Experts like Torres caution against excessive optimism, as unexpected hurdles and setbacks may arise. They point out that the field of AI has experienced periods of decline in the past, referred to as winters. During these periods, corporate interest and funding for AI diminished significantly. This historical context underscores the need for careful consideration of the risks and uncertainties associated with AI development.
As the debate over the future of AI continues, it is crucial to balance the optimism surrounding its potential benefits with a realistic evaluation of the risks involved. The insights and perspectives provided by AI researchers are crucial in navigating this complex landscape. By fostering ongoing dialogue and collaboration, society can make informed decisions about the role and impact of AI in our lives.
