The work of making synthetic intelligence that holds to the guardrails of human values, recognized within the business as alignment, has developed into its personal (considerably ambiguous) area of examine rife with coverage papers and benchmarks to rank fashions in opposition to one another.
However who aligns the alignment researchers?
Enter the Center for the Alignment of AI Alignment Centers, a company purporting to coordinate hundreds of AI alignment researchers into “one last AI middle singularity.”
At first look, CAAAC appears authentic. The aesthetics of the web site are cool and calming, with a brand of converging arrows paying homage to the thought of togetherness and units of parallel strains swirling behind black font.
However keep on the web page for 30 seconds and the swirls spell out “bullshit,” making a gift of that CAAAC is all one large joke. One second longer and also you’ll discover the hidden gems tucked away in each sentence and web page of the fantasy middle’s web site.
CAAAC launched Tuesday from the same team that introduced us The Box, a literal, bodily field that girls can put on on dates to keep away from the specter of their picture being become AI-generated deepfake slop.
“This web site is crucial factor that anybody will examine AI on this millenium or the subsequent,” stated CAAAC cofounder Louis Barclay, staying in character when speaking to The Verge. (The second founding father of CAAAC wished to stay nameless, in line with Barclay.)
CAAAC’s vibe is so much like AI alignment analysis labs — who’re featured on the web site’s homepage with working hyperlinks to their very own web sites — that even these within the know initially thought it was actual, together with Kendra Albert, a machine studying researcher and expertise lawyer, who spoke with The Verge.
CAAAC makes enjoyable of the development, in line with Albert, of those that wish to make AI protected drifting away from the “actual issues taking place in the actual world” — equivalent to bias in fashions, exacerbating the vitality disaster, or changing staff — to the “very, very theoretical” dangers of AI taking on the world, Albert stated in an interview with The Verge.
To repair the “AI alignment alignment disaster,” CAAAC shall be recruiting its international workforce solely from the Bay Space. All are welcome to use, “so long as you imagine AGI will annihilate all people within the subsequent six months,” in line with the roles web page.
Those that are keen to take the dive to work with CAAAC — the web site urges all readers to carry their very own moist gear — want solely touch upon the LinkedIn publish asserting the middle to robotically turn into a fellow. CAAAC additionally presents a generative AI device to create your individual AI middle, full with an govt director, in “lower than a minute, zero AI data required.”
The extra formidable job seeker making use of to the “AI Alignment Alignment Alignment Researcher” place will, after clicking via the web site, finally discover themselves serenaded by Rick Astley’s “By no means Gonna Give You Up.”