AI forecast scenario

[terrafamilia]

There's a new forecasting scenario spearheaded by former OpenAI researcher Daniel Kokotajlo.

https://ai-2027.com/

This scenario posits all-around superhuman AI capabilities by the end of 2027. 
I am not qualified to determine its credibility, I'm just irked by the use of "compute" as a noun. 

Ciao,

Terrafamilia

[Drashner1]

I skimmed this but didn't read it all the way through. It's interesting, but not particularly new by OA standards other than the more recent timeframe.

My main issue with it is the apparent blind spots in the article. All this superhuman intelligence and technological advancement but corporations are still the primary vehicle of economic activity, job losses are a problem of the common people while CEOs and software people stay on (at least for a while - why a superhuman AI needs them around is unclear), and people deprived of the 'benefits' of spending the majority of their time working to further the dreams of corporate shareholders/CEOs/the neo-liberal consensus cease to have any meaning in their lives and sink into hedonistic do-nothingism. Bleh.

Todd

[Worldtree]

this timeline only goes up to october 2027 , so i can see why the scenarios depicted here are fairly conservative predictions. I wouldn't expect a radically different world in terms of social structures and institutions

this is entirely about the near future and projecting the effects of ChatGPT-like software only 2-3 years from now seems like a reasonable excercise?

there's probably a few ... political instabilities it's glossed over, but that's to be expected with a neural scenario like this.
the graphic design is nice too https://www.lightconeinfrastructure.com/

[ProxCenBound]

I saw this a while back and skimmed it. It's interesting, though I'd bet that when 2028 rolls around, the situation won't really be much different than it is now. I think there needs to be more theoretical breakthroughs, and that current methods aren't on track for that sort of takeoff yet.