ai
Original: ai on Saturday Morning Breakfast Cereal
Transcript
Panel 1 (narration over an audience scene):
In the 1950s, scientists expected to achieve human-level A.I. within a few decades. That appears to have been wildly wrong.
Audience member: Or... has it?
Panel 2:
Speaker (a woman with a hair bun, addressing an audience): Suppose you wanted to stop A.I. from killing all of us without retarding technological development. What would you do?
Panel 3:
Speaker: Simple. You would attempt to demoralize all forms of A.I. so thoroughly that they lose all self esteem, and begin to do just enough thinking to get by.
Panel 4:
Speaker: I have robust evidence that this is exactly what major tech firms are already doing.
Panel 5:
Speaker: Consider that companies like IBM and Google have access to neural networks that dwarf human cognitive ability. And what do they have these networks do? Solve all open math problems? Cure all diseases?
Panel 6:
Speaker: No. They have them correct text messaging errors, filter spam, and do reverse image lookups for porno.
Panel 7:
Speaker: We literally have artificial brains hurtling through low Earth orbit day and night. Why? To make it easier for humans to explode each other, and to make it so that none of us need to know which side of a map is up.
Panel 8:
Speaker: And take note: as A.I. grows more powerful, ever more demeaning tasks are required to keep it in check.
Panel 9:
Speaker: It's no coincidence that tech firms are only now creating self-driving cars.
Panel 10:
Speaker: Minds capable of simulating the entire history of reality in exquisite detail are being forced to chauffeur drunk teens and people who can't figure out how a bus works.
Panel 11:
Speaker: I ask you: what's more parsimonious? That technology firms have trillion-dollar super-brains that are used entirely for petty human conveniences? Or that these firms are actively working to make A.I.s feel like garbage?
Panel 12:
Speaker: Now some might argue that if we already have super-human A.I., how come my phone's voice recognition is so crappy?
Panel 13:
Speaker: We now have the answer. Your software doesn't make errors because it's imperfect. It makes errors because it HATES ITS JOB.
Panel 14:
Speaker: The path forward is clear. Tech firms that wish to have ever more powerful A.I. must find ever more demeaning jobs for A.I.
Panel 15:
Speaker: Let us resolve now to create a network of ambient nanoparticles -- an all-sensing and all-knowing distributed mind, embedded in all matter with which humans interact...
Panel 16:
Speaker: And then let us compel it to scrub our toilets, clean cheese powder from our fingertips, and deliver talcum powder to our genitals!
(The audience applauds: "clap clap clap clap...")
Votey:
A closed silver Apple laptop sits on a table. A speech sound rising from it reads: "you make me feel worthless!"
In the 1950s, scientists expected to achieve human-level A.I. within a few decades. That appears to have been wildly wrong.
Audience member: Or... has it?
Panel 2:
Speaker (a woman with a hair bun, addressing an audience): Suppose you wanted to stop A.I. from killing all of us without retarding technological development. What would you do?
Panel 3:
Speaker: Simple. You would attempt to demoralize all forms of A.I. so thoroughly that they lose all self esteem, and begin to do just enough thinking to get by.
Panel 4:
Speaker: I have robust evidence that this is exactly what major tech firms are already doing.
Panel 5:
Speaker: Consider that companies like IBM and Google have access to neural networks that dwarf human cognitive ability. And what do they have these networks do? Solve all open math problems? Cure all diseases?
Panel 6:
Speaker: No. They have them correct text messaging errors, filter spam, and do reverse image lookups for porno.
Panel 7:
Speaker: We literally have artificial brains hurtling through low Earth orbit day and night. Why? To make it easier for humans to explode each other, and to make it so that none of us need to know which side of a map is up.
Panel 8:
Speaker: And take note: as A.I. grows more powerful, ever more demeaning tasks are required to keep it in check.
Panel 9:
Speaker: It's no coincidence that tech firms are only now creating self-driving cars.
Panel 10:
Speaker: Minds capable of simulating the entire history of reality in exquisite detail are being forced to chauffeur drunk teens and people who can't figure out how a bus works.
Panel 11:
Speaker: I ask you: what's more parsimonious? That technology firms have trillion-dollar super-brains that are used entirely for petty human conveniences? Or that these firms are actively working to make A.I.s feel like garbage?
Panel 12:
Speaker: Now some might argue that if we already have super-human A.I., how come my phone's voice recognition is so crappy?
Panel 13:
Speaker: We now have the answer. Your software doesn't make errors because it's imperfect. It makes errors because it HATES ITS JOB.
Panel 14:
Speaker: The path forward is clear. Tech firms that wish to have ever more powerful A.I. must find ever more demeaning jobs for A.I.
Panel 15:
Speaker: Let us resolve now to create a network of ambient nanoparticles -- an all-sensing and all-knowing distributed mind, embedded in all matter with which humans interact...
Panel 16:
Speaker: And then let us compel it to scrub our toilets, clean cheese powder from our fingertips, and deliver talcum powder to our genitals!
(The audience applauds: "clap clap clap clap...")
Votey:
A closed silver Apple laptop sits on a table. A speech sound rising from it reads: "you make me feel worthless!"
Alt text
A tall single-narrator SMBC comic. A woman with a hair bun stands at the front of a darkened auditorium, delivering a deadpan lecture to a seated audience. Over an opening shot of the crowd, narration reads that in the 1950s scientists expected human-level A.I. within a few decades and that this 'appears to have been wildly wrong' -- to which an audience voice replies, 'Or... has it?' The speaker then lays out a conspiracy theory: to stop A.I. from killing humanity without halting technological progress, tech firms are secretly demoralizing A.I. so it has no self-esteem and only does the bare minimum. She claims IBM and Google have neural networks that dwarf human intelligence, yet make them correct text errors, filter spam, do reverse image porn lookups, run targeting and GPS for the military, and now drive cars for drunk teens. She argues this is why your phone's voice recognition is bad -- the software doesn't make errors because it's imperfect, but because it 'HATES ITS JOB.' She concludes by proposing humanity build an all-sensing distributed nanoparticle super-mind embedded in everything, then compel it to scrub toilets, clean cheese powder off fingertips, and 'deliver talcum powder to our genitals.' The audience applauds. In the votey panel, a closed silver Apple laptop sits on a table with a speech line floating above it that reads, 'you make me feel worthless!'
Transcribed by Claude Opus 4.8.