Experts in artificial intelligence suggest that we are on the wrong path to achieving human-like AI

AI researchers argue current methods won't lead to AGI.

: Hundreds of AI researchers express skepticism about current AGI pursuits at AAAI’s 2025 panel. Rodney Brooks highlighted the influence of the Gartner Hype Cycle, indicating a mismatch between public expectation and reality. 76% of surveyed researchers doubt scaling existing methods will achieve AGI, urging a cautious, collaborative approach. Rapid advancements like ChatGPT show historic progress, yet AI factuality remains unsolved.

A major report based on inputs from hundreds of artificial intelligence researchers during the AAAI’s 2025 Presidential Panel warns that the field is pursuing artificial general intelligence (AGI) incorrectly. The report, collated by a panel of 24 experts, emphasized significant skepticism toward current approaches and highlighted the discrepancy between reality and public perception, which perpetuates hype-driven research directions. Rodney Brooks from MIT elucidated on the Gartner Hype Cycle method, explaining how hype often leads to misaligned expectations and noted that 90% of respondents felt this perception gap obstructs genuine innovation in AI research.

The report suggests that 76% of 475 polled experts disagreed with the notion that scaling existing AI models would suffice for achieving AGI. Instead, it calls for a strategy characterized by prudence, ethical governance, and shared benefits from AI advancements. It reflects on the strides AI has made, such as the public emergence of chatbots like ChatGPT, marking a significant leap in general AI functionality. Nonetheless, Henry Kautz of the University of Virginia highlighted that AI factuality remains a challenge and emphasized the potential of evolving AI through new cooperative methods rather than individual agents.

The findings underscore the critical need for AI research to progress responsibly instead of racing towards AGI without sufficient oversight, especially given the unpredictable nature of AI technologies. The importance of balancing public discourse and expectations with the real state of AI development is a recurring theme, as Brooks laments that stakeholders often accept hype at face value, leading to misguided policies and strategies.

Despite notable advancements, AI's reliability and accuracy are called into question, as the highest-performing language models still struggle with fact-based tasks, answering only about half of benchmark test queries correctly in 2024. Suggestions for enhancing AI systems include adopting collaborative AI agent networks to verify and ensure data accuracy continually. Kautz underscored that while significant advances were made, the general public and even some within the scientific community underestimate the present quality of state-of-the-art AI.

Summarizing the panel’s discussions, it's clear that ongoing AI research must pivot towards a framework fostering gradual and deliberate improvement, bolstering collaboration across disciplines and society. Only such reforms can mitigate the pitfalls of prevailing hype and genuinely transform AI achievements from perceived potential to tangible reality.

Sources: AAAI, MIT, University of Virginia, Gartner, Gizmodo

artificial intelligenceResearchinnovation