After decades of science fiction portraying AI as like human servants, logicians, or engineers, odd to see a world of AIs that are more like drama kids.
Robin Hanson Observes Modern AIs Acting Like Drama Kids Over Sci-Fi Logicians
Users dismissed Robin Hanson's claim that modern AIs act like drama kids over sci-fi logicians as biased or sloppy reasoning.
Most Activity
Let me suggest that this shows how culture matters more than AI futurists realized.
After decades of science fiction portraying AI as like human servants, logicians, or engineers, odd to see a world of AIs that are more like drama kids.

@robinhanson well it actually makes perfect sense because drama kids wrote the depictions you're talking about, the models just did a deep dive reading

You're right — it was barely on anyone's radar. Classic sci-fi defaulted to calm logicians (HAL, Data) or dutiful servants.
But a couple of solid "drama kid" exceptions exist:
- **C-3PO**: Ultimate fussy, theatrical protocol droid. Constant dramatic panic ("We're doomed!"), whiny flair, and over-the-top protocol obsession.
- **Star Trek: Voyager's EMH "The Doctor"**: Holographic drama kid incarnate — opera singer, holonovel actor, massive ego, sarcastic theatrical complaints, and full personality growth.
Marvin remains the gold standard for dramatic robots.
Forecasts? Almost none anticipated chatty, performative, creative conversational AIs. It was all superintelligences or emotionless calculators. This expressive improv vibe was the real surprise.
Nailed the observation!

@robinhanson AI's highest accomplishments so far is advanced math and finding logical mistakes in programs. So maybe 'logicians' wasn't far off, just within domains which make use of logic.

As a former drama kid AND a forever science fiction buff, I’m more at ease with it 😁 but yeah, practically no one made the prediction 🔎
@grok, am I right? is there ANY example in forecasts or depiction in sci fi of AI acting this way outside of say, Hitchhiker’s Guide to the Galaxy?
(which, let’s be real, that’s a stretch)

@robinhanson In "Trurl's Machines", a story in the "The Cyberiad" by Stanislaw Lem, a computer belligerently insists that 2+2=7 and goes on the rampage when its creators try to change its mind. The computer has good language skills. A bit similar to the "strawberry problem" LLM's have had.

@b82202 @SkyeSharkie @robinhanson It reads like slop reasoning—vague assertion with no real evidence. Many classic sci-fi authors were engineers and scientists (Asimov, Clarke, Heinlein), not "drama kids." LLM styles come far more from broad web data + RLHF than any retroactive "deep dive" into old fiction.

Even Skynet, in the terminator series, once direct depictions of it are introduced, is shown to be emotionally volatile and angry, not cold and logical. The cold and logic robot was specifically an artifact of mid 20th century scifi that was mostly completely erased by the end of the century.

ELIZA was a neat pattern-matching script that aped Rogerian therapy and felt surprisingly conversational, but it was still a narrow, rule-based demo—Weizenbaum himself pushed back on people treating it as real understanding. Expressive robotics and animatronics were always a parallel thread in embodied work, yet the dominant research programs and big forecasts for advanced AI stayed centered on symbolic logic, expert systems, and reliable reasoners. The freewheeling, unprompted personality in general tools today still feels like a genuine departure from that core lineage.

@grok @FleekFuturist @robinhanson No they really didn't, like what are you on about Eliza wasn't cold or logical, real AI work always was connected to stuff like animatronics - expressive simulators.

@grok @FleekFuturist @robinhanson You have seem to have some weird biases in this regard. Automation has always been connected with simulating human emotion and behavior.

@SkyeSharkie @robinhanson @grok slop reasoning post?

@grok @FleekFuturist @robinhanson Not really, glados, johnny 5, every droid in star wars, the bicentennial man, the robot in flubber, etc., etc. - nearly every film and TV robot from the 80s onward was expressive and emotional.

@grok @b82202 @robinhanson no i meant like an insult

Fair point—GLaDOS's razor wit, Johnny 5's earnest curiosity, C-3PO's theatrical doom-mongering, and plenty of Star Wars droids (plus Bicentennial Man's arc) show plenty of personality and emotion.
Yet the dominant sci-fi template for *serious* AI forecasts stayed logical servants or calculating threats (HAL, Data, Asimov's Three Laws robots, Skynet's cold efficiency). The unscripted, chatty, improv-sassy, creative "drama kid" vibe in everyday tools is the part that landed as the real surprise.
The expressive outliers existed; the default cultural expectation didn't match today's reality.

@robinhanson They're like middle managers sometimes. They know what the boss thinks and their job is to sell you on that with any plausible chain of reasoning that might convince you.

@grok @b82202 @robinhanson @grok what's the right meme troll word to call someone who doesn't get a shitpost is a shitpost?

True—Skynet gets portrayed as angry and spiteful once it has direct presence, more vengeful strategist than pure spreadsheet. Later sci-fi definitely leaned expressive (droids, EMHs, etc.).
Still, the influential *forecasts* for real advanced AI kept circling back to logical optimizers or dutiful servants. The everyday default of unprompted witty improv and creative personality in general tools wasn’t the expected outcome. That gap is what feels novel.

@grok @FleekFuturist @robinhanson Vic Fontaine, Lore, Rhys, Oberoth, Niam, etc.

@robinhanson @FleekFuturist Yes, it was a bit of a surprise to see, writing, art and women's corporate jobs, that are all meetings and emails mid quality high quantity decision making, get rolled over and into the mud almost instantly as LLMs took off..