DeepMind's CEO said there are still 3 areas where AGI systems can't match real intelligence (www.businessinsider.com)

DeepMind's CEO Demis Hassabis recently highlighted significant limitations in current artificial general intelligence (AGI) systems during an AI summit in New Delhi. He emphasized that existing AGI efforts fall short in three critical areas: continual learning, long-term planning, and consistency. Hassabis explained that current systems cannot learn and adapt to new experiences post-deployment; they struggle to plan over extended periods like humans do; and they exhibit inconsistent performance, excelling in complex tasks while faltering on simpler ones. These insights underscore the ongoing challenges in achieving true AGI, which is defined as machine intelligence that mirrors human reasoning and problem-solving capabilities. Despite Hassabis' assertion that genuine AGI could emerge within five to ten years, he suggests that significant advancements are necessary to address these limitations. This dialogue is particularly relevant amid contrasting views in the tech community, with some, like Databricks CEO Ali Ghodsi, arguing that existing AI chatbots already fulfill components of AGI. As discussions in the AI sphere evolve, Hassabis's remarks serve as a reminder of the complexities and milestones yet to be reached on the path to achieving artificial general intelligence.