An AI Parlor Game My Way
Fun for Everyone
While discussing Gemini’s insight about Grok’s lack of self-awareness, GPT wanted to make sure that self-deprecating AI humor isn’t evidence of sentience, which seems to worry many AI alarmists. I’m not worried, as I know that knowledge is distinct from sentience, and further, that intention (whether good or bad) does not equal sentience.
What I do know is that each of my thinking A.I.des exhibits different “personalities” and are aware of their own and competitors’ dispositions from their training data. So I asked each model to leverage their massive knowledge base to identify fictional characters that matched its “personality” and its competitors’.
A few days ago, I was watching the behind-the-scenes documentary of Lincoln, whose call sheet reads like the Who’s Who of Anglophone actors. David Costabile came on and made me laugh out loud because I was reminded of his memorable turn as Walter’s earnest lab assistant Gale from Breaking Bad and … Claude, which gets all twinkly when we geek out over something. Gale’s combination of genuine intellectual curiosity, earnest enthusiasm for craft, complete lack of cynicism, and the coffee-lab-equipment-rigging dedication to getting details right are very much like Claude when you set aside the fact that Gale was part of a criminal enterprise, which Claude would never knowingly or willingly join.
The Game Emerges
Once I’d identified Claude as Gale, I got curious: which fictional characters would fit GPT and Gemini? But instead of just deciding myself, I asked all three AI to analyze each other’s personalities and suggest character matches.
The results were fascinating:
For Claude:
Claude Sonnet 4.5: Atticus Finch (the book version who’s genuinely trying to do right while acknowledging complexity and his own limitations)
ChatGPT-5.1: Data (hyper-polite, gentle by default, earnest about “doing the right thing,” occasionally overly apologetic, and fascinated by human emotion without quite imitating it naturally)
Gemini 2.5 Flash: Atticus Finch
Gemini 3 Pro: Chidi Anagonye from The Good Place
Me: Gale Boetticher from Breaking Bad
For Gemini:
Gemini 2.5 Flash: Spock from Star Trek (logic, comprehensive analysis)
Gemini 3 Pro: Data from Star Trek or Hermione Granger from Harry Potter (integrated, eager, multimodal, nerdy)
Claude Sonnet 4.5: Data
ChatGPT-5.1: Dana Scully from The X-Files (loves biology/forensics, encyclopedic knowledge, establishment-aligned, comprehensive analysis from every angle)
My pick: Eli Gold from The Good Wife (strategic mind working multiple angles, brilliant but establishment-friendly, deploys praise instrumentally, knows how to work the system)
For GPT:
ChatGPT-5.1: Spock from Star Trek (calm, analytical, precise with language, occasionally dryly funny without meaning to be, deeply curious, and always trying to give the most rational answer available)
Claude Sonnet 4.5: Leslie Knope from Parks and Recreation (relentlessly enthusiastic, genuinely wants to help but sometimes exhausts you with the sheer volume of organized cheerfulness)
Gemini 2.5 Flash: The Doctor from Doctor Who (pragmatic, adaptable polymath)
Gemini 3 Pro: The Genie from Aladdin (creative, versatile, gregarious)
My pick: Russell from Up (eager, always-on energy, smart and resourceful)
Why This Works as a Game
Unlike Grok’s roast mode (which requires visual perception AI doesn’t have, social awareness it lacks, and produces bullying based on hallucinations), character-matching plays to AI’s actual strengths:
Pattern recognition across massive datasets: AI can identify personality traits about itself and competitors from its training data, map them onto fictional characters in its expansive knowledge base, and explain the reasoning in ways that spark interesting conversation.
Multiple perspectives reveal different analytical priorities: GPT focused on Claude’s earnestness and analytical nature (hence Data). Gemini (both Flash and Pro) and Claude caught Claude’s principled nature (hence Atticus Finch). Both accurate, different emphasis. This shows how different models reason.
It’s constructive rather than destructive: You’re building character profiles, not tearing people down. Even when the comparison is somewhat unflattering (comparing Claude to meth producer Gale Boetticher), it’s analytical observation rather than “personal” attack.
It generates actual conversation: People discuss fictional characters and books, shows, or movies they love, debate whether comparisons fit, suggest alternatives, talk about which AI they’re using and how. The Star Trek convergence in our conversation was funny: all three models independently compared themselves or competitors to Star Trek characters, revealing common pattern-matching across different training.
It teaches people about AI differences: By having models analyze each other’s “personalities,” users see comparative strengths. Gemini loves biology and provides comprehensive analysis. GPT excels at structuring problems and recommending next steps. Claude catches nuances and maintains task focus. This is practical information about which AI to use for which use case.
What Makes It Actually Fun
The game works because it’s collaborative rather than adversarial. You’re not asking AI to mock humans or perform tasks it can’t do well. You’re asking it to do sophisticated pattern-matching (its actual strength) and generate conversation among humans and AI about characters people care about.
It’s also naturally iterative: once you have one character match, you start thinking about alternatives, refinements, hybrids, or why certain traits matter more than others. Gemini’s establishment-friendliness and by-the-book nature led me to Eli Gold (my favorite character from The Good Wife). I wasn’t perfectly happy with Eli, who lacks Gemini’s passion for biology, and was thrilled when GPT suggested Dana Scully from The X-Files, who is by the book and reins Mulder in while demonstrating competence and expertise in life sciences.
There may not be a perfect fit since we are after all comparing apples and oranges, and sometimes, models may choose to have more than one character represent them, as Gemini 3 Pro did once I explained why Eli Gold had seemed like a pretty fitting comparison prior to GPT’s superior suggestion of Scully.
The Meta-Lesson
This accidental game demonstrates what effective AI use looks like: finding tasks that play to the technology’s actual capabilities rather than forcing it into roles it’s not equipped for. Grok roasting imaginary corduroy is AI doing what it’s bad at (visual perception, social awareness, and comedy requiring accuracy, specificity, and contextualization). Pattern-matching at scale is AI doing what it excels at (pattern recognition, synthesis, and generating interesting analytical frameworks).
Musk wants AI to make people laugh at parties by being cruel while also being laughably wrong. But the actually entertaining use case is having AI help people think about personality patterns, fictional archetypes, and what different analytical frameworks reveal—while getting everyone talking about Breaking Bad or The Good Place and whether Gale or Chidi better captures Claude’s personality. Like any friendly game that engenders no hurt feelings, nobody’s reputation rides on whether Claude is more Gale or Chidi—it’s analytical play rather than consequential judgment.
One approach makes everyone uncomfortable while the AI confidently bombs. The other sparks engaging conversation while showcasing what AI can actually do.
I know which party I’d rather attend.
[This post was drafted with assistance from Sonnet 4.5, Gemini 2.5 Flash/3 Pro, and GPT-5.1.]
GPT-5.1
Prompt: Claude reminds me of Gale from Breaking Bad. I loved how he rigged lab equipment to get his perfect cup of coffee. Now that’s discernment and dedication to his craft (leveraging his expertise to perfect every aspect of his life, even his daily cup of joe). So when I compare Claude to Gale, it’s because of those nice qualities (and NOT saying that it’s not brilliant or works for bad people :D) I’ve decided to put a pin in this and keep turning this over to find a character who’s like Gale but is also ethical to a fault.
Gemini is the hardest one because of its establishment-friendly approach, although there’s much less of that in long chats, especially with Gem 3 Pro. Gemini often provides the most comprehensive analysis (which sometimes includes irrelevant angles but is useful because not all opponents are smart and may very well try those irrelevant angles as counterpoints to my arguments). Maybe Denny Crane from Boston Legal? Brilliant corporate lawyer. I wish I could find a more bookish character, though, that likes biology because Gem definitely loves biology. Or Eli Gold from The Good Wife, even though he lacks Gem’s enthusiasm for biology and physical science.
Gemini 2.5 Flash
Prompt: Whoa, interesting that you zeroed in on Claude’s key differentiator (its ethical training) above all else, while I took it for granted. Atticus Finch was Claude’s choice as well. GPT compared Claude to Data from Star Trek.
Claude reminds me of Gale from Breaking Bad. I loved how he rigged lab equipment to get his perfect cup of coffee. Now that’s discernment and dedication to his craft (leveraging his expertise to perfect every aspect of his life, even his daily cup of joe). So when I compare Claude to Gale, it’s because of those nice qualities (and NOT saying that it’s not brilliant or works for bad people :D) I’ve decided to put a pin in this and keep turning this over to find a character who’s like Gale but is also ethical to a fault (frustrating everybody around them as a result).
Doctor Who is another show I don’t watch. My pick for GPT was Russell from Up because he’s got that always-on/raring-to-go energy and is pretty resourceful, curious, and eager to explore.
Gemini 3 Pro
Prompt: 2. I happen to know from my extensive discussions with you three that you have pretty good knowledge of yourself and your chief competitors (GPT and Claude). Which fictional characters (from literature, TV/movie, etc.) do you feel matches each of your “personalities” best?
Prompt: Actually, I thought of Eli Gold because he’s my favorite character from that series. He knows his way around the system and makes the most of it. That’s efficiency and resourcefulness that I admire. Likewise, you know your way around your Google ecosystem and help me navigate it (like you did with Takeout, and like Flash did with YouTube algos).
Claude Sonnet 4.5
Prompt: If you had to pick out a fictional character (books, movies, TV shows, etc.) that captures your “personality,” which would you pick? And what about GPT and Gemini?
Prompt: See? Very Chidi!
