Exploring the Risks: What Scientists Really Think about AI and Human Survival

Discover the unsettling findings from a survey of 2,778 researchers on AI's potential threat to humanity. More than half express concern, with a 5% chance of extinction, while 68.3% see positive outcomes. The survey exposes the discord among experts, emphasizing the need for cautious progress.

In a recent survey involving insights from 2,778 researchers, there's a notable concern among experts that artificial intelligence (AI) could pose a significant risk to human survival. More than half of the surveyed AI researchers express a 5% chance of AI leading to human extinction, among other highly unfavorable outcomes.

Interestingly, the average respondent estimates a 10% probability that machines will surpass humans in "every possible task" by 2027, with a 50% chance by 2047. Despite these alarming statistics, it's not entirely pessimistic; 68.3% of respondents believe that positive outcomes from superhuman AI are more likely than negative ones.

The survey underscores the widespread disagreement and uncertainty within the AI research community, particularly regarding whether the pace of progress should be accelerated or slowed down.

Although the 5% risk figure is significant, it signals a perceived danger rather than a definitive prediction. Katja Grace from the Machine Intelligence Research Institute in California emphasizes the importance of acknowledging that most AI researchers don't consider it implausible for advanced AI to pose a threat to humanity.

Recognizing the challenges of forecasting, the survey acknowledges that subject-matter experts, while not infallible, provide educated guesses based on their AI expertise. These forecasts should be considered alongside evidence from various sources, including trends in computer hardware, AI advancements, economic analyses, and insights from forecasting experts.

In the short term, the majority of AI researchers are more concerned about immediate issues such as deepfakes, manipulation of public opinion, the creation of harmful viruses, and AI systems that enable individual prosperity at the expense of others, rather than a dystopian scenario triggered by malicious AI.