AI progress is surprisingly revealing all over the world, a recent survey reveals.
While looking at the survey responses from 2778 AI researchers who met the authors’ criteria for expertise and comparing them to the previous survey, the AI researchers discovered that, overall AI experts perceive a sense of speeding up of progress across the board.
To explain it in simpler terms, the average time frame for AI progress in different niches like Microsoft, Windows, and Apple moved up by a year between the two surveys, showing a significant change in experts’ expectations regarding the acceleration of AI development.
The survey’s basic findings include a remarkable shift in collective forecasts for two main concepts: High-Level Machine Intelligence (HLMI) and Full Automation of Labor (FAOL) compared to similar forecasts made in 2022. For HLMI, the estimated arrival time dropped by 13 years between 2022 and 2023, while the forecast for FAOL decreased by a substantial 48 years over the same period.
This survey shows a significant change in how the experts view things than normal people. The AI experts have concluded that “for any kind of task, mechanics can perform better and more quickly than human workers” The ultimate output will be almost 50 years sooner than they previously believed.
Considering how fast the experts think these material consequences will arrive. It’s interesting to see what they think about the speed at which AI should develop. Some so-called “effective accelerationists” believe that it should happen faster while others, AI “doomers” think it should happen slower. But the AI doomers who want AI progress “much slower” were the smallest, at 4.8 percent, and the ones who are in favor were much more significant at 15.6 percent, outnumbering the doomers.
When asked about the speed at which AI should develop, the group of people who wanted the “somewhat slower” was the largest, with 29.9 percent of responses. This was followed by “current speed” at 26.9 percent and “somewhat faster” at 22. 8 percent. Essentially, the majority of responses, accounting for 79.6 percent, leaned towards maintaining the current pace or making only slight adjustments.
It is quite important to note an important detail mentioned in the survey: the respondents are experts in AI, not specifically in forecasting, and may lack expertise in non-technical factors influencing AI’s trajectory. This cautionary note is relevant whenever reading about AI experts predicting the future.
However, these findings are not significant just because AI researchers can’t predict the future with quite accuracy. These people play an important role in advancing technology, and understanding their subjective beliefs provides insights into what they normally want, fear, and anticipate in their fields. So, they think that an AI-driven automated world will arrive sooner than expected, and as a collective, they are somewhat undecided about whether the pace of AI progresss is positive.