At the India AI Impact Summit 2026, Google AI CEO Demis Hassabis emphasized the limitations of AI systems like Gemini and ChatGPT, particularly in solving complex Maths Olympiad problems. While these systems demonstrate advanced reasoning skills, they frequently falter on basic mathematics when questions are posed differently. Hassabis highlighted that true general intelligence should exhibit consistent performance and continual learning capabilities. Current AI models, he noted, are “frozen” post-training, lacking the ability to adapt to new experiences or develop long-term strategies. He called for advancements that enable continuous learning and coherent planning across tasks. In a broader context, Hassabis sees significant potential for AI in healthcare, particularly through drug discovery innovations at Isomorphic Labs. He predicts that AI could revolutionize the development of new medicines, potentially finding solutions to all diseases over the next two decades, thus underscoring the transformative role of AI in enhancing human health.
Source link
