Synthetic intelligence (AI) security NGOs have proliferated in recent times, particularly within the US and EU. These organizations — akin to FAR AI, the Way forward for Life Institute, the Machine Intelligence Analysis Institute, and the Middle for AI Security — declare to defend the general public from AI dangers. Their focus typically consists of excessive long-term eventualities like an AI “singularity” (a hypothetical level the place AI surpasses human intelligence and escapes human management). On this publish, we’ll study who these teams are, the place their funding and affect come from, and the way credible their claims of rogue AI fashions “escaping” actually are. We’ll break down the AI takeover danger by mannequin kind — particularly giant AI fashions edging towards common intelligence — and assess the technical feasibility of an AI spreading itself between machines. Alongside the way in which, we’ll have a look at some exhausting numbers (with sources) on mannequin sizes, computational necessities, and comparisons to conventional laptop viruses. Lastly, we’ll think about how a lot public cash goes into these AI security efforts and whether or not their affect is warranted or simply fear-mongering.
