Last week, on the decaying husk of X.com, a former Disney Channel actor posted a 90-second advertisement for his AI company, 2WAI. The ad centers around “Baby Charlie” interacting with the avatar of his deceased grandmother.
In scene 1, baby Charlie’s pregnant mother asks avatar-grandma for advice. In scene 2, she asks avatar-grandma to tell toddler Charlie a bedtime story. In scene 3, 10-year-old Charlie talks with avatar-grandma while walking home from school. In scene 4, 30-year-old Charlie shows avatar-grandma an ultrasound photo, and she reacts with delight. And then, in scene 5, we see Charlie’s mom recording a three-minute video with still-alive grandma, to upload for the avatar. Then the tagline: “with 2wai, three minutes can last forever.”
I’ve written before about the social role of technology (here, here, and here. It’ll also be in the book). Advertisements like this are meant to build hype and garner attention, but they also present a vision of how a new technology fits into our social world.
The 2wai commercial isn’t in the same genre as Sam Altman and Dario Amodei heralding the coming age of Artificial Superintelligence. The product they are promising is basically feasible with existing generative AI technologies. (Sora 2, etc) But the social vision is fucking ghastly.
There is a potential near-future where you could conceivably record a video of your parent and then offload much of the day-to-day interactions with your child to an LLM wearing their face and voice as a digital skinsuit. This was basically the premise for a Black Mirror episode (“Be Right Back,” S2 E1). Black Mirror is not a fucking Pinterest board, former Disney Channel actor! Your ideas are derivative and bad. Have better ideas.
Season 4, Episode 5 of Black Mirror (“Metalhead”) takes place in a near-future dystopia where the scattered remnants of humanity hide from robot dogs that kill anyone they encounter. It is a dark tale.
WIRED’s Will Knight reported last week that researchers at Anthropic have taken steps to go build the damn thing.
“We have the suspicion that the next step for AI models is to start reaching out into the world and affecting the world more broadly,” Logan Graham, a member of Anthropic’s red team, which studies models for potential risks, tells WIRED. “This will really require models to interface more with robots.”
Anthropic, of course, continues to brand itself as the good guys in the AI race. They aren’t trying to build an AI-enabled robot army. They’re just trying to see if they can, before someone else1 builds the AI-enabled robot army.
And look, I continue to think that Eliezer Yudkowsky’s AI Doomer argument in If Anyone Builds It, Everyone Dies is weak and riddled with holes. I am not particularly worried that a runaway superintelligence is going to extinguish life on earth anytime soon.
But if you hook flawed LLM technologies up to robots, and then Silicon Valley defense tech companies arm the robots… a lot of things can very predictably go wrong.
This is bad and I hate it. The point of Black Mirror is to critique the current trajectory of how society makes use of new technology. The point of Black Mirror isn’t to give ghoulish AI entrepreneurs a fucking dream journal.
