AGI could be achieved in next 5-8 years: Google DeepMind
What's the story
Demis Hassabis, the CEO of Google DeepMind, has said that Artificial General Intelligence (AGI) is closer than ever and could be achieved within the next five to eight years. Speaking at the AI Impact Summit in New Delhi, Hassabis stressed the need to address biosecurity and cybersecurity risks as AI systems continue to evolve. "We're starting to see what these systems can do," he said.
Risk mitigation
Need to address biosecurity, cybersecurity risks
Hassabis stressed the importance of ensuring that defenses against AI risks are stronger than the offenses. He said, "I think we need to worry about things like bio and cyber risk in AI very soon. The current systems are getting pretty good at cyber." This highlights his concern over potential threats posed by advanced AI systems and their cybersecurity capabilities.
System shortcomings
Continuous learning gap in AI systems
Hassabis, who created AlphaFold, pointed out that one of the biggest gaps in today's AI systems is their inability to learn continuously. "Today, systems are trained, fine-tuned and then essentially frozen before being released into the world," he explained.
Future prospects
AI will boost productivity of scientists, experts
Looking ahead, Hassabis said AI will greatly boost the productivity of scientists and experts. He said cross-disciplinary research that requires understanding multiple fields at once will benefit the most from AI. This is because it can help researchers find connections that are otherwise hard to spot.