Luo Yonghao's judgment: comedy is the last text category AI will conquer
He polled top-tier large language models on text creation and got consistent answers: stand-up comedy, with its one-liner density—text so tightly packed with jokes—can theoretically be conquered by AI, but will inevitably rank last among all text types. Which means comedy writers will outlive other creators by several years in the age of AI.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
Excessive fear kept our ancestors alive
When early humans heard rustling in the grass, those who overestimated danger—treating every sound as a predator—survived; the relaxed ones got eaten. Modern people inherited this hypervigilance; hormonal shifts trigger baseless alarms everywhere. Men especially suffer in silence, feeling emasculated by admitting fear.
Learning to refuse unreasonable requests is safer than fearing conflict
A listener was pestered by a street vendor after an open-mic show. Too afraid to refuse, she paid repeatedly—then the vendor chased after her claiming she hadn't paid. Two comedians had to rescue her. One pulled her aside afterward: ‘Next time, refuse.’ She realized that fearing confrontation more than consequences is what puts you in danger.
You don't have to become outgoing to accept connection
A guest advised a socially anxious listener: don't force yourself to become social; just muster courage to not reject when others invite you or show affection. He shared his own experience: he used to decline 100% of social advances. He changed only that one thing—and everything got easier. You don't have to rewire your whole life; just practice catching what others throw.
AI invents legal citations that don't actually exist
An intellectual property court worker encountered AI-generated briefs that looked airtight at first glance—even lawyers initially found them persuasive. But scrutiny revealed invented case numbers that don't exist, and citations of laws unrelated to the actual case. Most people have no way to catch these fabrications with their own knowledge.
Cross-verify between three top models to catch AI errors
For work that cannot afford mistakes, Luo's solution: subscribe to the three or four best models globally. Take the first result and paste it verbatim into the second, third, and fourth—let each search online for latest facts and catch errors. He uses this constantly and hasn't seen an error survive three independent checks. Caveat: only works with the very top tiers, not the budget models.
— Luo YonghaoComedy ranks last in what AI will conquer
When asked about his fear of AI, Luo gave a specific judgment: he's asked top models about text creation and got consistent answers—stand-up comedy, with its densely-packed one-liners, is a problem AI will theoretically solve, but will definitely come last among all text types. So comedians will outlast all other writers. And comedy requires live audience testing; AI can't fake that in the near term.
— Luo YonghaoEveryone researches AI; he still edits scripts by hand
He confesses he feels no change from the AI wave—everyone around him is debating and diving deep, while he still edits comedy the stone-age way: no phone calls with comics, no video. He has to show up in person and sit face-to-face for forty-fifty minutes to hash out one script. Recently it's been relentless. He jokes he works like a caveman, and sometimes that dumb method brings odd reassurance.
— Luo YonghaoAI remembered a shopping suggestion from two years ago
A listener shared an uncanny experience: two years earlier she'd asked AI for advice on buying a certain product, then didn't purchase. Months later, shopping for the same category, AI unprompted mentioned that old recommendation was actually superior, now discontinued, and suggested finding it secondhand—better than anything new on the market. She'd completely forgotten the conversation; checking chat history confirmed every word. She found unopened items listed on the secondhand platform.
In their own words · checked verbatim
So I asked her, I was asking her, I said, ‘Baby, which floor are you going to?’ Nobody spoke. I don't know why nobody spoke. My classmate wasn't speaking either.
然后我就跟她 我就问她嘛 我就说宝宝你去几楼呀 没人说话 我不知道为什么没人说话 我同学也不说话
Then Doubao (豆包) gives you an answer that looks really logical. Sometimes what people upload makes even us lawyers think, ‘Hmm, that's actually got some merit.’ But when you look closer, you find things in there like ‘case number 12345’—which doesn't exist at all.
然后豆包就会给他一个 看似非常有逻辑的答案 甚至有的时候 他们传上来的东西就是会让我们法律从业者都会觉得 好像有点道理 但是你细看就会发现里面会出现 比如说判决书12345 就这个是根本不存在的
Get subscriptions to the three or four best models. When you're doing serious work, take what the first one outputs, paste the whole thing into the second, third, and fourth ones, and let them find errors.
你买最好的三到四家大模型 然后你做严肃工作的时候 用第一个出来结果以后 你把它全文导入给第二第三第四个 让他挑错
Luo Yonghao2:08:48
Comedy is the last text genre that will be conquered. I've asked several of the top models about this, and they all reached the same conclusion: comedy text creation—especially stand-up, with one punchline per line, that high-density writing style—is something AI will theoretically eventually overcome, but will definitely rank last. Which means comedians in the AI age will have several more years to live.
文本的写作里最后被淘汰的是喜剧 因为对AI来讲 就我问了好几个最Top的大模型 都是一个结论 就是喜剧文本的创作 像尤其脱口秀这种一句一个梗 这种高密度的这种文本写作 是AI最终才能理论上能克服 但是一定排在最后 所以对喜剧工作者来讲 在AI时代是能多活几年的
Luo Yonghao2:10:48
My work is like a caveman's. Like a cook over a wok. Right now if a comic's script needs revision, I have to discuss how to fix it—I can't even call or video chat. I have to physically go there.
我的工作就像个原始人 就跟炒菜厨子一样 就是我现在跟一个演员 比如他稿子要跟他聊聊怎么改的问题 我甚至都不能跟他电话聊或者视频聊 我都得人肉过去
Luo Yonghao2:10:48
So I think if you haven't experienced that flow state with AI, that feeling of it communicating with you from the soul level—if you haven't gotten that feeling, then you're probably either not using it right, or not using it enough.
所以我觉得 如果大家没有跟AI聊出那种心流的感觉 自己的感觉 感觉他从灵魂深处能跟你进行交流 这个感觉如果没聊出来 大概率就是还用的不太对 或者是用的不够多
Even people working in AI globally right now—one hand fear, the other excitement. It points to this: you get joy, happiness, that deep satisfaction of solving unknowns from this work. But there's a shadow hanging over it—it might wipe out humanity. Both are coexisting now.
就是现在全世界从事AI工作的人 也是一边是恐惧 一边是兴奋 最终指向是 就你有很多快乐幸福和那种 解决未知问题的那种满足感来自这儿 但也有悬着一个阴影 就是他可能消灭人类 这两个现在是并存的
Figures
| AI-generated news coverage in short-form video | 50%+ | 2:14:54 |
How to listen
Listeners who enjoy Luo Yonghao and stand-up comedy scene gossip, and want to hear how comedy professionals genuinely assess AI's impact on creative work.
The first two hours are fear-game improvisation largely unrelated to AI; skip directly to 2:02:46.