Nobel laureate Demis Hassabis recently spoke to the Financial Times about the role and impact of AI in scientific research. He believes that AI has reached a watershed moment, where people recognize that it is mature enough to aid in scientific discovery.
Hassabis stated that AlphaFold heralds a new era of scientific discovery. He hopes that in a decade's time, looking back at AlphaFold, it will signify the beginning of a new golden age for scientific discoveries across various fields.
Image source: The image was generated by AI, provided by the image licensing service Midjourney
He also discussed how AI is a new tool for scientific research. AI can help scientists solve long-standing problems, such as the protein folding issue. Hassabis noted that AI is an engineering science, which means you must first create AI before you can study it.
Hassabis also emphasized the importance of understanding AI. He believes that understanding is a crucial part of the scientific method. AI is a man-made, engineered artifact, and this does not mean its complexity is less than the natural phenomena we wish to study.
He further mentioned that achieving AGI requires more time, effort, and thought to be invested in understanding and analyzing tools, benchmarks, and evaluations. Hassabis reiterated that understanding is a crucial part of the scientific method. Ideally, this understanding means you make fewer mistakes.
Hassabis also pointed out that achieving AGI involves addressing a social issue: what values do these systems possess? What goals do we want to set for them? He believes the second question might be more challenging to answer, such as defining what the goals and values are.