Time and again, experts in the field of artificial intelligence have raised their concerns about how AI, in general, could pose dangers to humanity. In particular, artificial general intelligence (AGI) is seen as especially concerning, adding to this alarming narrative.
Now, Google DeepMind, arguably one of the biggest players in the generative AI space, has now come forward with a research paper that highlights these very same fears. The paper discusses how AGI poses significant risks—risks serious enough to cause “substantial harm.” But at the same time, it outlines various preventive measures as to how to safely work with it, utilising its benefits.
Also Read: Oppo Find X8s full specifications and images tipped online ahead of April 10 launch
Research Paper Outlines Preventive Measures
The research paper, titled An Approach to Technical AGI Safety and Security, delves into the transformative benefits of AGI, while also outlining the major risks it presents. These risks are categorised into four key areas: misuse, misalignment, mistakes, and structural risks.
Regarding the misuse category, Google states that its strategy involves preventing threat actors from accessing dangerous AI capabilities. This includes actively identifying such capabilities, implementing robust security measures, restricting access, continuous monitoring, and ensuring model safety.
For misalignment, Google outlines two primary lines of defence. The first is model-level mitigation, and the second involves system-level security measures.
Google also explains how all these measures can be combined to create comprehensive safety cases for AGI systems as a whole
“AI, and particularly AGI, will be a transformative technology. As with any transformative technology, AGI will provide significant benefits while posing significant risks. This includes risks of severe harm: incidents consequential enough to significantly harm humanity,” reads the research paper.
Google DeepMind’s CEO Has Previously Expressed Concerns, Too
This isn’t the first time that Google DeepMind has voiced concern about the fast-growing AI race. Previously, its CEO, Demis Hassabis, a Nobel winner, in an interview with Axios, said that we will see “powerful agentic systems,” and that they “are going to be built, because they'll be more useful, economically more useful, scientifically more useful. ... But then those systems become even more powerful in the wrong hands, too.”
{{/usCountry}}This isn’t the first time that Google DeepMind has voiced concern about the fast-growing AI race. Previously, its CEO, Demis Hassabis, a Nobel winner, in an interview with Axios, said that we will see “powerful agentic systems,” and that they “are going to be built, because they'll be more useful, economically more useful, scientifically more useful. ... But then those systems become even more powerful in the wrong hands, too.”
{{/usCountry}}He added, “I think there needs to be more time spent by economists, probably, and philosophers and social scientists on what do we want the world to be like, even if we get everything right, post AGI," "I'm surprised there's not very much discussion about that, given the relatively short timelines."