Is AGI the right goal for AI? (garymarcus.substack.com)

A recent Op-Ed in The New York Times questions whether pursuing Artificial General Intelligence (AGI) is the right objective for AI development at this time. The author argues that large language models (LLMs), while ambitious, face significant issues like hallucinations and inaccuracies, leading to disappointing returns on investment for companies. A study from MIT's NANDA Initiative revealed that a staggering 95% of organizations saw little to no value from their AI pilot projects. The piece contends that the tech industry should shift its focus from generalized AI tools to narrow, specialized applications that can effectively address specific challenges in fields like science and medicine. In a related publication, a collaborative paper involving prominent researchers, including Yoshua Bengio, offers a nuanced definition of AGI as "an AI that can match or exceed the cognitive versatility and proficiency of a well-educated adult." This revised perspective emphasizes the complexity of cognition, breaking it down into various facets rather than relying solely on economic criteria. While acknowledging that current metrics for assessing AI capabilities may not be comprehensive or enduring, the paper aims to spark ongoing dialogue regarding the fundamental nature of intelligence and the goals of AI research.