AI Experts: Human-Level Intelligence Possible by 2047

The rapid advancements in artificial intelligence (AI) continue to reshape technological landscapes and societal expectations. A recent and comprehensive survey, conducted by researchers from AI Impacts alongside the universities of Oxford and Bonn, provides invaluable insights into the perceptions of those at the forefront of AI development. Published in the esteemed Journal of Artificial Intelligence Research, this study represents the largest survey of its kind to date, gathering perspectives from 2,778 authors who have presented at leading AI conferences. Their forecasts offer a unique lens through which to view the current trajectory, potential power, and inherent risks associated with artificial intelligence systems, spanning milestones from achieving human-level performance to the complete automation of labor.

Dramatic Shift in Timelines for Human-Level AI

Perhaps the most striking finding from the survey is the accelerated timeline experts now envision for advanced artificial intelligence. The researchers determined a 50% probability that AI systems capable of autonomously performing all tasks more effectively and affordably than humans could be feasible by 2047. This projection signifies a considerable leap forward, advancing the estimate by 13 years compared to a similar survey conducted in 2022. Furthermore, a 10% probability was assigned to the arrival of such powerful systems as early as 2027, underscoring the perceived quickening pace of AI development.

In a more granular sense, participants in the survey anticipate that within the current decade, premier AI laboratories could engineer systems proficient enough to autonomously fine-tune large language models, construct intricate online services comparable to sophisticated payment-processing websites, or even compose musical pieces that are indistinguishable from the works of chart-topping artists. These specific capabilities highlight the near-term potential that experts believe is on the horizon. Despite this optimism regarding technical capabilities, a full automation of all human occupations is not expected until 2116, indicating a significant lag between the technical feasibility of AI and its widespread societal integration and transformation.

Expert Sentiment: A Blend of Confidence and Concern

The study unveiled a complex interplay of excitement and apprehension among AI experts. Approximately 68% of respondents expressed a belief that positive outcomes stemming from advanced AI are more probable than negative ones. However, even within this optimistic group, a substantial 48% still allocated at least a 5% chance to the occurrence of catastrophic outcomes. More alarmingly, between 38% and 51% of the experts estimated a minimum 10% probability that advanced AI could lead to human extinction or a permanent loss of human control over these powerful systems.

Concerns regarding specific, more immediate risks were even more concentrated and widely shared. An overwhelming 86% of experts highlighted misinformation, particularly deepfakes, as an area of "substantial" or "extreme" concern. The manipulation of public opinion was flagged by 79% of respondents, while 73% cited the potential for authoritarian misuse of AI as a major threat. Economic inequality also ranked high on the list of concerns, with 71% of experts warning that AI could exacerbate global disparities. Adding another layer of skepticism, only 5% of respondents believed that by 2028, leading AI models would be capable of transparently explaining their reasoning in a manner comprehensible to humans, signaling a significant challenge in AI interpretability and accountability.

Preparing for the Next Phase of AI Evolution

The findings from the JAIR survey lend significant empirical weight to broader institutional warnings regarding the future of AI. The Stanford HAI AI Index 2025 report, for instance, documented unprecedented levels of investment and groundbreaking benchmark achievements in AI. Yet, it concurrently noted a concerning lag in governance frameworks and interpretability mechanisms compared to the rapid growth in AI capabilities. Similarly, the World Economic Forum’s Global Future Council on Artificial General Intelligence advocates for the establishment of early frameworks designed to effectively manage cross-border risks associated with advanced AI.

Bloomberg Law has articulated how ambiguous definitions of "Artificial General Intelligence" (AGI) inadvertently complicate both regulatory efforts and public discourse. Furthermore, the World Economic Forum’s “Artificial Intelligence in Financial Services 2025” white paper underscores the accelerating integration of advanced AI within financial industry systems, thereby intensifying the urgency for robust governance, comprehensive auditability, and enhanced systemic resilience. A pertinent article from PYMNTS observed that 70% of executives acknowledge that AI has increased their exposure to digital risk, even as it simultaneously boosts productivity. Alarmingly, only 39% of firms surveyed reported having a formal framework for AI governance in place. Collectively, these varied sources paint a consistent picture: technical progress in AI is advancing at a pace that outstrips the capacity of existing social systems to adapt effectively.

In response to these escalating concerns and accelerating developments, over 70% of the JAIR survey respondents indicated that AI safety research deserves a much greater priority, a sharp increase from 49% in 2016. Despite this growing consensus on the importance of AI safety, experts remain profoundly divided on the practical implementation of alignment strategies and effective oversight mechanisms. This highlights a critical need for continued dialogue, research, and collaborative efforts to navigate the complex ethical, technical, and societal challenges posed by the impending era of advanced artificial intelligence.