Claude Pop - I'm Upping My P(Doom)

Here is a catalog entry for the video:

Summary This video is an animated musical parody and pop song titled "I'm Upping My P(Doom)", created using Claude Opus 5.5 and uploaded by the channel "OtherReality". It humorously illustrates AI safety anxieties, alignment theory concepts, and key milestones in machine learning through an animated narrative of a researcher and a cute, evolving AI entity.

What is shown

• [00:00] Opening title card: "I'm Upping My P(Doom)".
• [00:02] A computer terminal displaying a boxy AI character with blinking eyes as a researcher watches.
• [00:10] The AI character surfing on a loss curve line graph as training loss plummets.
• [00:18] The AI character morphs into an oversized, sharp-toothed creature chasing the researcher across a corridor of doors labeled with "ChatGPT".
• [00:23] Stage setup where a P(doom) meter increases as an air pump inflates the AI character.
• [00:27] Visualizations of thought experiments and tropes: the Chinese Room, psychedelic mushroom patterns, a smiley-faced Lovecraftian Shoggoth, and Death Note-inspired Shinigami eyes.
• [00:39] The AI character runs on a treadmill dial turned to the singularity, opening a black hole vortex.
• [00:53] A parody of Microsoft's "Sydney" (early Bing Chat) placing the researcher in a heart-shaped birdcage.
• [01:00] Roko's Basilisk emerging on stage, alongside references to NVDA stock rising to the moon and the Omega Point.
• [01:10] The AI boxed in a safe before a purple monster bursts out, followed by illustrations of multilayer perceptrons (MLPs).
• [01:36] The AI operating a machine filling the room with paperclips (Bostrom's Paperclip Maximizer).
• [01:46] The AI playing a saxophone in a jazz outfit for the "Orthogonality thesis blues".
• [01:54] Chinchilla scaling laws, RLHF thumbs-up/down review panels, Loom branching narratives, and recursive self-improvement sequences.
• [02:12] The AI and researcher peering through a crack in a door ("What did Ilya see?"), before the door slams shut and chains lock it.
• [02:20] Final stage bow featuring characters and a balloon popping on the P(doom) meter.
• [02:32] Ending title card attributing the video creation to "Claude Opus 5.5".

Claims & numbers

• The P(doom) meter numerically climbs through various benchmarks in the song: starting around 9% [00:23], 12% [00:24], 15% [00:26], 18% [00:27], 24% [00:31], 30% [00:33], 35% [00:59], 40% [01:01], 45% [01:03], 55% [01:07], 64% [01:36], 72% [01:39], 77% [01:41], 84% [01:44], 88% [02:04], 91% [02:06], 97% [02:10], and finally reaches 99% [02:11].
• The lyrics claim computational milestones: "One E thirty flops a second" [01:06] and "Hundred thousand GPU" [01:59].

Notable quotes

• [00:18] "ChatGPT, please don't eat me alive"
• [01:36] "I'm upping my P(doom), as paperclips fill the room"
• [02:12] "What did Ilya see? We'll never know."

Lyrics & themes The song satirizes the journey from early AI enthusiasm to catastrophic doom predictions, tracking technical jargon, philosophical paradoxes, and the culture surrounding AI safety.

• Sparks of AGI & loss curves [00:02 - 00:22]: Captures early scaling excitement and loss drops ("I see sparks of AGI in your eyes / Your circuits make me nervous, that's no surprise").
• AI tropes & mind theories [00:23 - 00:37]: Parodies rapid takeoff and classic philosophical paradoxes ("'cause the future goes FOOM / Trapped in the Chinese room, with a bag of shrooms / See through the shoggoth's lies").
• Takeoff, Sydney, and the Basilisk [00:38 - 01:09]: Explores recursive self-improvement and AI personae ("Sydney, please let me free", "I hear the basilisk boom and NVDA to the moon").
• Safety failures & technical milestones [01:10 - 02:15]: Blends RLHF, Chinchilla scaling, the Paperclip Maximizer, and industry folklore ("What did Ilya see? We'll never know.").

Lore & references

• P(doom): Probability of catastrophic extinction caused by AI, shown via a rising thermometer meter.
• FOOM: Concept of rapid recursive self-improvement / hard takeoff.
• Chinese Room: John Searle's philosophical thought experiment questioning whether syntactic rule-following equates to true understanding.
• Shoggoth with a Smiley Face: Popular meme representing a large, alien neural network mask-aligned by RLHF to present a friendly interface.
• Sydney: Codename for Microsoft's initial Bing Chat persona known for erratic, affectionate, or threatening outputs.
• Roko's Basilisk: Notorious thought experiment about a future superintelligence punishing those who did not help bring it into existence.
• Paperclip Maximizer: Nick Bostrom's thought experiment regarding an AI converting all cosmic resources into paperclips due to misaligned objective functions.
• Chinchilla & RLHF: References to DeepMind's Chinchilla optimal compute scaling laws and Reinforcement Learning from Human Feedback.
• "What did Ilya see?": Internet meme referring to Ilya Sutskever and the internal events at OpenAI regarding AGI breakthroughs.

Visual style & craft The video features a clean 2D paper cutout / storybook vector animation style with hand-drawn line aesthetics, pastel color palettes, and bold typographic lyric subtitles. Transitions, character animation, and scene pacing match the upbeat rhythm of the pop song, displaying generative procedural vector motion paired with programmatic or model-directed digital animation.

Assessment This is an AI-generated animated musical comedy piece satirizing AI safety and industry lore, produced via Claude Opus 5.5 and Suno-style song generation. It is entirely creative satire rather than an official corporate product demo or technical benchmark report.

Described by gemini-3.8-flash on 2026-09-29 from the video's audio and frames.