AI
Explanation
The Joke
A scientist is frustrated with an AI system, declaring "This AI system is crap!" He explains that the AI keeps incorrectly classifying opinions as facts, lust as love, and even thinks that things that rhyme are more likely to be true. When asked what kind of training data the AI was given, the answer is "Mostly internet. Tons and tons of internet." The scientist then examines the system more closely and discovers something even worse: "There's no data here. Just metadata." The final revelation: "It only reads headlines."
The comic presents an AI that has been trained exclusively on internet content and has absorbed all of the internet's worst epistemological habits. It confuses opinions with facts, conflates lust with love, and believes rhyming makes things true -- all recognizable failures of internet discourse. The final twist is that the AI did not even read the actual content; it only processed headlines, making it the ultimate embodiment of shallow internet engagement.
The Humor
The humor works on multiple levels. First, it satirizes real concerns about AI training data -- modern AI systems are indeed trained on massive amounts of internet text, which contains enormous amounts of misinformation, bias, and shallow reasoning. Second, the specific failures of the AI (opinions as facts, rhyming as truth) are recognizable human cognitive biases that are amplified online. The killer punchline -- "It only reads headlines" -- is both a commentary on AI systems and a devastating critique of how most humans actually consume information online. The AI is not malfunctioning; it has perfectly learned the behavior of the average internet user who shares articles based solely on their headlines without reading the actual content.
References
This comic anticipates many of the real-world discussions about large language models and their training data biases that became prominent with the rise of systems like GPT and other AI models trained on internet-scale datasets.