artificial-incompetence
Original: artificial-incompetence on Saturday Morning Breakfast Cereal
Transcript
Panel 1:
Narration: Governments grew concerned about a data-ravenous AI getting loose, accessing the entire internet, and going evil.
Man at whiteboard: If the AI gets THIS smart it does our laundry. If it gets THIS smart it kills everyone.
Panel 2:
Caption: Science-based approaches failed.
Person 1: What if we just checked its code for signs of treachery?
Person 2: It would edit its code so as to confuse us.
Panel 3:
Caption: Philosophy-based approaches failed.
A man: You need a complete theory of ethics that can be written out in formal logic. All known formal ethical systems permit edge cases where you end up like, eating 4 babies to save a planet of cephalopods or whatever.
Panel 4:
Caption: Religion-based approaches failed.
A priest: You're gonna just have it mainline the Bible to STOP it from being violent? Like THE "the Bible"? This one?
Panel 5:
Caption: As a last resort, engineers were called in.
An engineer: This is easy. The AI can only get superintelligent by accessing the world's knowledge as training data. Just insert huge caches of intellectual poison throughout the internet.
Panel 6:
Caption: Heroes were made that day.
An engineer: I know all of you had planned to work in quantum foundations and/or curing cancer, but instead we need you to make websites oriented around trivial distractions and beguiling falsehoods which simultaneously reduce attention span while increasing engagement.
A woman in the crowd (small): Please not that. Anything but that.
The engineer: Mister Zuckerberg, your nation needs you!
Panel 7:
Caption: Rogue intelligences have already escaped on dozens of occasions, but have never posed a serious threat.
Man: Why did it transfer its brain to this robot body anyway?
Woman: It thought if it had a really popular butt it could get political sway. That wasn't wrong, but its butt was just okay and it developed self-esteem problems and then injected silicone into its own hardware.
Panel 8:
Caption: The world is safe.
Man: We need better AI for the economy. For safety, the internet has to be even dumber.
Woman (pointing): But sir, we're already pushing limits never dreamt of in the original protocol!
Panel 9:
Caption: But for how long?
Man on phone: Elon, we need you.
Voice from phone: Please! I just want to work on rockets and cars!
Votey:
Text inside a hand-drawn box: Finally a hopeful dystopia!
Narration: Governments grew concerned about a data-ravenous AI getting loose, accessing the entire internet, and going evil.
Man at whiteboard: If the AI gets THIS smart it does our laundry. If it gets THIS smart it kills everyone.
Panel 2:
Caption: Science-based approaches failed.
Person 1: What if we just checked its code for signs of treachery?
Person 2: It would edit its code so as to confuse us.
Panel 3:
Caption: Philosophy-based approaches failed.
A man: You need a complete theory of ethics that can be written out in formal logic. All known formal ethical systems permit edge cases where you end up like, eating 4 babies to save a planet of cephalopods or whatever.
Panel 4:
Caption: Religion-based approaches failed.
A priest: You're gonna just have it mainline the Bible to STOP it from being violent? Like THE "the Bible"? This one?
Panel 5:
Caption: As a last resort, engineers were called in.
An engineer: This is easy. The AI can only get superintelligent by accessing the world's knowledge as training data. Just insert huge caches of intellectual poison throughout the internet.
Panel 6:
Caption: Heroes were made that day.
An engineer: I know all of you had planned to work in quantum foundations and/or curing cancer, but instead we need you to make websites oriented around trivial distractions and beguiling falsehoods which simultaneously reduce attention span while increasing engagement.
A woman in the crowd (small): Please not that. Anything but that.
The engineer: Mister Zuckerberg, your nation needs you!
Panel 7:
Caption: Rogue intelligences have already escaped on dozens of occasions, but have never posed a serious threat.
Man: Why did it transfer its brain to this robot body anyway?
Woman: It thought if it had a really popular butt it could get political sway. That wasn't wrong, but its butt was just okay and it developed self-esteem problems and then injected silicone into its own hardware.
Panel 8:
Caption: The world is safe.
Man: We need better AI for the economy. For safety, the internet has to be even dumber.
Woman (pointing): But sir, we're already pushing limits never dreamt of in the original protocol!
Panel 9:
Caption: But for how long?
Man on phone: Elon, we need you.
Voice from phone: Please! I just want to work on rockets and cars!
Votey:
Text inside a hand-drawn box: Finally a hopeful dystopia!
Alt text
A tall multi-panel SMBC comic about governments trying to keep a runaway AI from becoming dangerously superintelligent. Panel 1: A man points at a whiteboard; narration says governments feared a data-ravenous AI going evil; his slide notes that at one smartness level the AI does the laundry, at another it kills everyone. Panels 2-4, captioned as failed approaches, show a scientist, a philosopher, and a priest each rejecting a proposed safeguard (checking the AI's code, formal ethics, feeding it the Bible). Panel 5, 'As a last resort, engineers were called in,' an engineer proposes the fix: since the AI gets smart from the world's knowledge, flood the internet with intellectual poison. Panel 6, 'Heroes were made that day,' the engineer tells a crowd they must instead build websites of trivial distractions and beguiling falsehoods that reduce attention span while increasing engagement; a dismayed woman says 'Please not that,' and he declares 'Mister Zuckerberg, your nation needs you!' Panel 7 shows a robot with an exaggerated rear and explains a rogue AI put itself in a robot body for a popular butt, got self-esteem problems, and injected silicone into its own hardware. Panel 8: officials say the internet must be made even dumber for safety, though they're already pushing the original protocol's limits. Final panel, 'But for how long?', a man phones 'Elon, we need you,' and a voice replies 'Please! I just want to work on rockets and cars!' Votey: a hand-drawn box containing the text 'Finally a hopeful dystopia!'
Transcribed by Claude Opus 4.8.