Keep pulling the thread on Demis Hassabis.
Demis Hassabis predicts that Artificial General Intelligence (AGI) will be achieved around 2030.
DeepMind's AlphaFold solved the 50-year grand challenge of protein structure prediction in biology.
Demis Hassabis believes that solving continual learning, long-term reasoning, and memory will be required to achieve Artificial General Intelligence (AGI).
Demis Hassabis estimates that no more than one or two major conceptual breakthroughs are needed to achieve Artificial General Intelligence (AGI).
Google DeepMind's smaller "flash" models are reportedly 95% as capable as its frontier models while costing only one-tenth the price to run.
Gemini or related AI technology is being integrated into every Google product, including Google Maps and YouTube.
Google DeepMind operates under the assumption that the capabilities of its frontier models can be replicated in small, edge-device-capable models within six to twelve months.
Demis Hassabis asserts that the lack of continual learning is a key factor preventing current AI agents from being able to perform complete, end-to-end tasks autonomously.
Current AI models are capable of solving International Mathematical Olympiad (IMO) problems at a gold medal level.
Google DeepMind has made a strategic decision to open-source its edge models for platforms like Android and robotics because they are considered inherently vulnerable to reverse-engineering once deployed on devices.
Demis Hassabis predicts that a fully functional virtual simulation of a biological cell will be developed within approximately 10 years.
AlphaFold is now used by over three million researchers, representing nearly every biology researcher worldwide.