I’m founder of the Existential Risk Observatory, a nonprofit aiming to reduce xrisk by informing the public since 2021. We have published four TIME Ideas pieces (including the first one on xrisk ever) and about 35 other media pieces in six countries. We’re also doing research into AI xrisk comms, notably producing to my knowledge the first paper on the topic. Finally, we’re organizing events, coupling xrisk experts such as Bengio, Tegmark, Russell, etc. to leaders of the societal debate (incl. journalists from TIME, Economist, etc.) and policymakers.
First, I think you’re a bit too negative about online comms. Some Yud tweets, but also e.g. Lex Fridman xrisky interviews, actually have millions of views: that’s not a bubble anymore. I think online xrisk comms is firmly net positive, including AI Notkilleveryoneism Memes. Journalists are also on X.
But second, I definitely agree that there’s a huge opportunity informing the public about AI xrisk. We did some research on this (see paper above) and, perhaps unsurprising, an authority (leading AI prof) on a media channel people trust seems to work best. There’s also a clear link between length of the item and effect. I’d say: try to get Hinton, Bengio, and Russell in the biggest media possible, as much as possible, as long as possible (and expand: get other academics to be xrisk messengers as well). I think eg this item was great.
What also works really well: media moments. The FLI open letter and CAIS open statement created a ripple big enough to be reported by almost all media. Another example is the Nobel Prize of Hinton. Another easy one: Hinton updating his pdoom in an interview from 10% to 10-20%, that was news apparently. If anyone can create more of such moments: amazingly helpful!
All in all, I’d say the xrisk space is still unconvinced about getting the public involved. I think that’s a pity. I know projects that don’t get funded now, but could help spread awareness at scale. Re activism: I share your view that it won’t really work until the public is informed. However, I think groups like PauseAI are helpful in informing the public about xrisk, making them net positive too.
I’m founder of the Existential Risk Observatory, a nonprofit aiming to reduce xrisk by informing the public since 2021. We have published four TIME Ideas pieces (including the first one on xrisk ever) and about 35 other media pieces in six countries. We’re also doing research into AI xrisk comms, notably producing to my knowledge the first paper on the topic. Finally, we’re organizing events, coupling xrisk experts such as Bengio, Tegmark, Russell, etc. to leaders of the societal debate (incl. journalists from TIME, Economist, etc.) and policymakers.
First, I think you’re a bit too negative about online comms. Some Yud tweets, but also e.g. Lex Fridman xrisky interviews, actually have millions of views: that’s not a bubble anymore. I think online xrisk comms is firmly net positive, including AI Notkilleveryoneism Memes. Journalists are also on X.
But second, I definitely agree that there’s a huge opportunity informing the public about AI xrisk. We did some research on this (see paper above) and, perhaps unsurprising, an authority (leading AI prof) on a media channel people trust seems to work best. There’s also a clear link between length of the item and effect. I’d say: try to get Hinton, Bengio, and Russell in the biggest media possible, as much as possible, as long as possible (and expand: get other academics to be xrisk messengers as well). I think eg this item was great.
What also works really well: media moments. The FLI open letter and CAIS open statement created a ripple big enough to be reported by almost all media. Another example is the Nobel Prize of Hinton. Another easy one: Hinton updating his pdoom in an interview from 10% to 10-20%, that was news apparently. If anyone can create more of such moments: amazingly helpful!
All in all, I’d say the xrisk space is still unconvinced about getting the public involved. I think that’s a pity. I know projects that don’t get funded now, but could help spread awareness at scale. Re activism: I share your view that it won’t really work until the public is informed. However, I think groups like PauseAI are helpful in informing the public about xrisk, making them net positive too.