Written by 7:26 pm Academic

### AI Researchers Surprised by Last Year’s Remarkable Progress in AI Development

The new findings also suggest that “accelerationists” may outnumber “doomers.R…

It’s not merely a figment of your imagination. As per a recent study, AI researchers themselves are amazed by the swift progression of AI development.

A recent comprehensive study of AI experts revealed that the advancement of AI is rapidly gaining momentum, particularly from the perspective of those within the field. This study sheds light on the significant gap between staunch AI proponents and AI skeptics, often referred to as AI “doomers,” who advocate caution due to concerns about potential AI-related catastrophes.

Despite this disparity, it seems that the die-hard AI enthusiasts may have the upper hand, subtly indicated in the survey results.

The preliminary findings of the survey were released by AI Impacts, a research firm based in San Francisco and funded by Open Philanthropy, spearheaded by tycoon and Twitter co-founder Dustin Moskovitz.

The researchers observed a unanimous sense of acceleration among AI experts by analyzing responses from 2,778 distinguished AI professionals who met the study’s stringent criteria. They noted that the expected timeframe for the feasibility of 32 different AI-related tasks shifted 1.0 times earlier between 2022 and 2023.

In simpler terms, the average projection for AI advancement moved up by a year between the 2022 study and the current one in 2023. This adjustment is more substantial than a mere affirmative response to the question, “Do you believe that progress in the field of Artificial Intelligence is accelerating?” It signifies that the experts revised their time estimates for progress on an annual basis.

The most significant revelations from the study pertain to the stark disparities in the experts’ forecasts regarding two critical concepts—High-Level Machine Intelligence (HLMI) and Full Automation of Labor (FAOL)—between 2022 and 2023. The estimated arrival time for HLMI decreased by 13-fold, while FAOL estimates plummeted by 48 times during this period.

This shift in perspective is truly remarkable. Experts now predict that the point at which machines could outperform human workers in any task may arrive almost 50 years earlier than previously anticipated, reflecting a remarkable change in outlook within just one year.

Considering the rapid pace at which these material advancements are expected, it is intriguing to delve into the experts’ opinions on the potential speed of AI development, contrasting the views of “effective accelerationists” against AI “doomers.” While only a small fraction (4.8%) identified as extreme doomers desiring a significantly slower AI progress, the majority favored a somewhat slower pace (29.9%), followed by those content with the current speed (26.9%), and those advocating for a somewhat faster development (22.8%). The accelerationists, envisioning a much faster AI evolution, comprised 15.6% of the respondents.

It is crucial to note a key distinction in the study: respondents possess expertise in AI rather than forecasting, either in general or specifically related to AI. Therefore, the authors suggest that these experts may have insights into non-technical factors influencing AI trajectory. This cautionary note emphasizes the importance of considering the limitations of predictions made by AI experts.

While AI researchers may not possess clairvoyant abilities, their insights remain valuable. These individuals are at the forefront of AI innovation, and their perceptions offer a glimpse into their hopes, fears, and visions for the future of AI. While they acknowledge the rapid emergence of an AI-driven automated world, there is a general uncertainty among them regarding the desirability of the pace of AI advancement.

Nevertheless, there is a growing cohort of individuals who express concerns about the accelerating pace of this technological evolution.

Visited 1 times, 1 visit(s) today
Last modified: January 12, 2024
Close Search Window
Close