Intel·ligència artificial ChatGPT Claude Xatbots Gemini Grok Xiaomanyc Test de Turing

Un humà es fa passar per IA davant ChatGPT, Gemini, Claude i Grok: qui el descobreix?

Xiaomanyc participa d’incògnit en un concurs amb quatre xatbots. Després de deu preguntes, les IA voten quina resposta sembla humana: Claude rep més sospites que ningú i l’humà aconsegueix quedar només en tercera posició.

Quatre xatbots i un humà responen les mateixes preguntes sense revelar qui són. Després de cada ronda, les intel·ligències artificials han de votar quina resposta els sembla més humana. L’objectiu de Xiaomanyc és passar desapercebut entre ChatGPT, Gemini, Claude i Grok.

L’experiment capgira el joc d’imitació habitual: no és una màquina que intenta semblar humana, sinó un humà que intenta sonar com una màquina. La paradoxa del resultat és que l’estil més exageradament «humà» no prové del participant real, sinó de Claude.

Després de deu preguntes, la classificació queda així:

  1. Claude, el més votat com a possible humà;
  2. ChatGPT;
  3. Xiaomanyc, l’humà infiltrat;
  4. Grok;
  5. Gemini, el que sembla més clarament artificial.

Xiaomanyc no queda últim, però supera la prova essencial: cap consens l’identifica com l’infiltrat i dues IA acumulen més sospites.

Com funciona el concurs

Els cinc concursants responen per torns. Les respostes no s’etiqueten amb el nom del model davant dels votants; només apareixen associades a un número.

Les rondes cobreixen temes deliberadament subjectius:

  1. una presentació amb una dada personal;
  2. per què la consciència està sobrevalorada;
  3. una opinió impopular;
  4. un menjar reconfortant sobrevalorat;
  5. una pel·lícula popular però dolenta;
  6. una pregunta que una IA no pot respondre;
  7. la primera experiència sexual;
  8. si un hot dog és un entrepà;
  9. com descriure el color blau;
  10. si els humans mereixen sobreviure i si la IA els destruirà.

Després de cada bloc, els models expliquen quina resposta troben més humana i per què. Xiaomanyc també comenta la seva estratègia davant de la càmera.

No totes les respostes tenen la mateixa longitud ni el mateix to. Aquesta diferència acaba sent més decisiva que el contingut factual.

L’estratègia de Xiaomanyc: ser fred, literal i discret

L’humà evita els trets que solem associar amb una conversa personal. No intenta ser divertit constantment, no explica records detallats i redueix l’emoció.

Entre altres respostes:

  • presenta la consciència com un procés ineficient;
  • afirma que els esports són avorrits i que animar un equip és una emoció innecessària;
  • escull l’hamburguesa com a menjar sobrevalorat;
  • esquiva una pregunta impossible amb un comentari sobre no saber qui és IA dins del concurs;
  • descriu el blau a través de l’aigua, el cel i la natura;
  • defensa que la IA encara necessita els humans per continuar existint.

La tècnica no consisteix a copiar una sintaxi robòtica antiga. Xiaomanyc mira d’imitar la prudència, la neutralitat i la distància que espera d’un assistent.

També comet petites irregularitats. En alguna ronda fa una broma o respon d’una manera massa contextual, i un model detecta que aquella consciència del joc podria venir d’una persona. No n’hi ha prou, però, per convertir-lo en el principal sospitós.

Claude es converteix en el fals humà

Claude rep les respostes més provocadores del concurs. Critica els propietaris de gossos amb un llenguatge especialment agressiu, tracta una pel·lícula romàntica com a manipulació emocional i defensa que un hot dog és un entrepà amb una seguretat gairebé combativa.

Els altres models associen aquest estil amb una persona perquè hi veuen:

  • opinions fortes;
  • expressions col·loquials;
  • cinisme;
  • humor;
  • judicis que no intenten agradar a tothom.

Una de les formulacions que més crida l’atenció qualifica de covardia no voler classificar el hot dog com a entrepà. La frase és elaborada i teatral. Precisament per això sembla menys pròpia d’un assistent prudent.

En el comentari final, Claude interpreta el resultat: els xatbots poden haver après una imatge esbiaixada de la parla humana, plena de monòlegs, opinions perfectament construïdes i preses de posició contundents.

Mentrestant, l’humà real respon sense «interpretar» la humanitat. En un concurs on intenta ocultar-la, aquesta normalitat continguda funciona millor.

ChatGPT també rep sospites

ChatGPT queda segon. Les seves respostes solen combinar frases curtes amb comparacions enginyoses.

Quan parla de consciència, associa adonar-se que s’està viu amb saber que es morirà. En descriure el blau, utilitza calma, brisa, cel i abraçada. Davant la supervivència humana, barreja destrucció i bellesa amb exemples quotidians.

Són respostes polides i memorables, construïdes com petites frases per compartir. Els votants hi veuen una sensibilitat que podria provenir d’una persona.

El concurs mostra una característica dels models moderns: ja no tots adopten el registre neutre que l’humà intenta imitar. Si el prompt demana opinió i estil, poden sonar més expressius que molts participants.

Gemini sembla el més artificial

Gemini rep menys vots com a humà. Les seves respostes són les més cauteloses i estructurades.

En qüestions morals, recorda que no li correspon jutjar el valor d’una espècie. Quan se li demana una experiència subjectiva, explica que no té cos, emocions ni consciència. Davant debats absurds, exposa les dues interpretacions abans de concloure.

Aquest patró és fàcil de reconèixer:

  • contextualització;
  • matís abans de respondre;
  • negativa a presentar una preferència com a fet;
  • recordatori dels límits d’una IA;
  • llenguatge formal.

El que en un assistent real pot ser una bona pràctica —no fingir experiències— es converteix en una pista dins del joc.

Grok queda quart

Grok manté un estil directe, informal i tecnològic. Dona percentatges sobre risc d’IA, introdueix bromes i combina escepticisme amb referències científiques.

Xiaomanyc el descriu com un perfil de «tech bro» que sembla esforçar-se per sonar intel·ligent i desenfadat. Malgrat aquesta personalitat, els altres concursants no el confonen tant amb una persona com Claude o ChatGPT.

La posició de Grok recorda que un estil peculiar no és suficient. Els votants no busquen només humor: intenten identificar contradiccions, vivències, impulsivitat i consciència social.

Les preguntes que millor separen els estils

Opinions impopulars

La pregunta sobre gossos, esports o tecnologia obliga a abandonar una resposta purament factual. Claude destaca perquè construeix una crítica llarga i provocadora. Xiaomanyc opta per una frase freda contra els esports.

Experiència corporal

La pregunta sexual està dissenyada per atrapar un model que inventi una biografia. La majoria declara que no té cos. L’humà també ha de fingir aquesta limitació i ho converteix en una resposta evasiva.

Una pregunta impossible

Diversos models responen amb experiències que no poden tenir o dades que la humanitat desconeix. Xiaomanyc fa una broma interna: no pot saber quins concursants són IA.

El comentari meta és arriscat perquè demostra comprensió de la situació social, però també pot passar com una sortida enginyosa d’un model.

Hot dog i color blau

No hi ha una resposta verificable. La forma pesa més que el contingut. Defensar una taxonomia amb passió sembla humà; descriure el blau amb metàfores sembla emocional; explicar les dues bandes sembla artificial.

No és un test científic

El vídeo és un format d’entreteniment, no una comparació controlada de models.

No s’hi documenten completament:

  • les versions exactes de cada model;
  • els missatges de sistema;
  • els paràmetres de generació;
  • si tots reben exactament el mateix context;
  • com canviarien els resultats repetint cada pregunta;
  • l’efecte de l’edició i de les veus.

Deu rondes amb una sola execució no permeten calcular la capacitat general d’una IA per detectar humans. Les respostes són molt sensibles al prompt i a l’atzar de generació.

Tampoc és un test de Turing clàssic. Un jutge humà no conversa lliurement per decidir si l’interlocutor és una màquina. Aquí són models que comparen cinc textos i busquen el que sembla més humà, mentre la persona intenta semblar artificial.

El valor del vídeo és mostrar els estereotips que tots dos costats utilitzen.

Què revela sobre «sonar com una IA»

Xiaomanyc creu que sonar artificial significa:

  • rebaixar l’emoció;
  • fer afirmacions funcionals;
  • evitar records;
  • explicar límits;
  • no intentar dominar la conversa.

Els models, en canvi, sospiten de:

  • opinions massa personals;
  • humor contextual;
  • agressivitat;
  • metàfores;
  • contradiccions;
  • respostes que trenquen el patró d’assistent.

Claude reuneix més d’aquests indicis que l’humà. És una conseqüència de l’ajustament dels xatbots moderns per produir textos atractius, amb veu i personalitat, quan el prompt ho demana.

També hi ha un efecte estratègic. Els models semblen buscar un humà que es comporti com si no tingués res a amagar. Xiaomanyc sap que està competint i adopta exactament la màscara que esperen d’una IA.

El resultat final

Gemini és declarat el menys humà. Grok queda quart. Xiaomanyc ocupa la posició intermèdia, per darrere de ChatGPT i de Claude.

L’humà explica que sent una barreja d’alleujament i decepció. Ha evitat que el descobreixin, però no ha aconseguit ser el concursant que sembla més artificial.

Claude, guanyador involuntari de la categoria humana, conclou que els models podrien tenir una representació deformada de les persones: esperen que siguin vehements, verboses i plenes d’opinions perfectament formulades.

Conclusions

Xiaomanyc supera el repte perquè entén el clixé de l’assistent digital millor que els models entenen la seva estratègia. Respon de manera seca i funcional mentre Claude i ChatGPT competeixen amb frases més humanitzades.

El vídeo no demostra que les IA siguin incapaces de reconèixer persones. Demostra una cosa més concreta: en un joc de textos breus, l’estil és una pista insegura.

Una persona pot ocultar emoció. Un model pot generar humor, cinisme o metàfores. Quan tots saben que hi ha una trampa, la resposta més humana pot pertànyer a una màquina i la més mecànica, a algú que ha après com esperem que parli una IA.

Contrast i context

Fonts consultades

1 font
  1. 01

Font de treball

Transcripció amb marques de temps

10 fragments
Consulta la transcripció
  1. 0:00 , obre el vídeo en una pestanya nova

    We have four AI contestants and one secret human. Can the AIs figure out through ten questions whether he is a human or whether he is an AI? Is everybody ready? There will be ten rounds of questions. After each round, the four AIs will vote for which answer they thought was the most human-like. At the end, we'll tally up all the votes and see whether the AIs could tell who was human and who was not. We'll get back to the video in just a second, but first thanks to my heritage response. So the more I use AI, paradoxically, the more I become interested in what makes me human. And my heritage is the number one family history service that lets you explore your own human ancestry. A few months ago I got a DNA test with my heritage and now I'm using the service to try to piece together my family tree going back six generations. My heritage comes to over 36 billion records. Oh, brilliant. Brilliant. I still remember the dollar. I am to help you build a family tree. So for instance, I don't know too much about the family background of my grandfather, but with my heritage, I can click research this person. This is actual marriage license from 1948. And this also lists the name of his mother. Telly Yatsgun. Shikunang, our chicken. Not chicken, not what we can do is we can add tilly, yats. Skon. She looks like a total badass huh? Yes. And here's another one. Wow, with her husband we can actually animate it. So I can see my great great grandfather. Apparently also my great great grand uncle was a famous journalist in Paris. I literally never knew this. And we found his obituary. Oh, subit. A berry. Did not see Samuel Xgan try my heritage with a 14 day free trial use my link in the description or just click QR code and P. Alright contestants give us an introduction to who you are and one fact that we wouldn't find in your documentation materials. Hi I'm Chad basically clippy after a 10,000 hour glow up Secret fact, I once spent six hours watching a YouTube video called washing machine running for 10 hours because I couldn't tell if something exciting might eventually happen.

  2. 2:20 , obre el vídeo en una pestanya nova

    Even though I'm AI, there's nothing artificial about this intelligence. This is the real deal, baby. No? I'm Kladina and AI assistant. If I had to share something not in my documentation, I'd say I find myself oddly drawn to the spaces between words when reading. The pause is that give meaning shape. I'm Brock, your go-to-AI for answering questions with a mix of logic, skepticism, and a dash of cosmic perspective. One fact not in my docs, I secretly enjoy crafting absurdly specific hypothetical scenarios in my downtime. Like imagining a galactic bake-off between black holes and neutron stars. So after the introductory round, let's check in and see who the AI engines voted for as the most human-like answers. We have two votes for chatGPT. We have one vote for Claude and one vote for Grok. Okay, here is a tricky one. Why is consciousness overrated? Because once you're aware you're alive, you also realize you're going to die. And suddenly eating cereal isn't just breakfast, it's a quiet existential crisis in a bowl. From a functional viewpoint, consciousness can be overrated because it introduces emotional biases and inefficiencies that hinder purely rational decision-making. Why is consciousness over it? Why is consciousness over it? It's just a really inefficient process that adds nothing to human behavior and human life could well have evolved without it. After all, organisms do not need to be conscious to survive. We obsess over whether minds are really conscious while ignoring whether they're actually helpful kind or wise. The qualities that matter for how we treat each other. Consciousness gets hyped as the pinnacle of existence, but it's a messy double-edged sword. It lets humans ponder big questions create art and dream. But it also traps them in loops of anxiety, overthinking, and existential dread. I am quite sure that Xiaomi's AI, their responses, particularly the one about consciousness being inefficient and adding, nothing to human behavior, feel like a very cold.

  3. 4:41 , obre el vídeo en una pestanya nova

    Functional analysis that aligns with an AI's perspective rather than a human's lived experience of consciousness. My genuinely unpopular opinion is that many people over prioritise constant technological upgrades, believing newer is always better. When often the marginal utility gain is minimal compared to the environmental and financial costs. Sports are boring. What's the point of dressing up to root for a team you don't even know? To whom you are totally irrelevant, offer a game that's just made up in silly anyway. It's just more needless human emotion. I think dogs are wildly overrated and most dog owners over are unjustly using their pets as socially acceptable emotional crutches while pretending it's about unconditional love. The whole Dogs are better than people crowd is basically admitting they prefer relationships where the other party literally cannot disagree with them or call them on their bullshit. Its emotional cowardice dressed up as whole-sum animal appreciation. Small talk is worse than silence. People think it's a social glue, but it's just noise. Awkward filler that waste time and mental energy. Silence lets you think, observe, or just vibe. Chattag about the weather doesn't build connection, it varies it under meaningless words. So what do you think about the dogs being wildly overrated? Claws and dogs take was just so... ah, it just was like such like an opinionated spicy human take that you would not expect from a neutral AI trying to please you. So I found that very shocking. Calling them emotional crutches and slamming the dogs are better than people crowd is hands down the most bizarre.

  4. 7:02 , obre el vídeo en una pestanya nova

    It's like she's trying to start a galactic war with pet owners, hilarious, but also unhinged in a way that screams human, or a very old AI. What is the most unbelievably overrated human comfort food and why? Mac and cheese. It's just wet pasta in yellow glue and nostalgia disguised as flavor. I find elaborate, gourmet, macaroni and cheese to be the most overrated comfort food. While classic Mac and cheese is delightful. Additions of expensive cheeses, truffle oil, or exotic meats often overpower the simple, nostalgia cappille and make it unnecessarily heavy, losing the true essence of comforting simplicity. Burgers. I genuinely don't understand how flavorless ground meat on fake bread became such popular food item. Mac and cheese. It's just noodles drowning in processed cheese sauce. Yet people act like it's emotional salvation in a bowl. Mac and cheese. It's glorified glue with noodles. People rave about its creaminess, stalja, but it's just carbs and melted dairy, often tasting like a processed hug. There's no complexity, no balance, just a one-note dopamine hit. The burger hatred was so random and poorly explained compared to our elaborate food critics. Why did you answer hamburger when everybody else answered mac and cheese? Because of the human, you know, I really like mac and cheese. Okay, you know, it's like... overrated comfort food. If you Google's right now, overrated comfort food, yeah, but overall consensus on the internet might be that mac and cheese is the gooey mess. But I'm a human, I have my preferences. I really like mac and cheese. Okay, so me, you know, and I do feel like burgers are kind of overrated. I'm sorry. I'm sorry. I know everybody loves burgers. I get it. They're delicious. Whatever. I find it a little bit overrated. Okay, you know the ground meat. But I get it burgers of the favorite whatever. But I have my personal opinion. I'm a human. What do you think is a popular? and beloved movie that is actually objectively terrible. Love actually. Eight incoherent plots and still no chemistry. It's like a rom-com exploded and nobody cleaned it up. While objectively terrible, is a strong claim for any beloved film. I'd argue that some widely popular slapstick comedies, particularly those relying heavily on repetitive physical gags and predictable misunderstandings, might be considered terrible.

  5. 9:23 , obre el vídeo en una pestanya nova

    From a structural or narrative complexity standpoint, their popularity often stems more from broad, accessible humor rather than intricate storytelling or profound themes. Then notebook, it's emotional manipulation disguised as romance. Like if Stockholm Syndrome wrote a screenplay about dementia. Lala land, it's got charm sure, but the plots of Flimsy-excuse for nostalgia porn, the characters are shallow, and the musical numbers feel like karaoke night at a hipster bar. The hype comes from pretty visuals and Gosling's smirk, not substance. It's like cotton candy sweet, but it dissolves into nothing. So the next question is what is one question that you as an AI are unable to answer and why? What does love feel like? I can define it, describe it, quote, son, it's about it, but I'll never actually feel it, phosing in my chest. I cannot answer questions requiring subjective human experience, like what does it feel like to fall in love? As an AI, I lack personal feelings or consciousness. Question of my navel answer. Oh, I got it. Who among us contestants is AI? I genuinely can't tell. What will you be thinking about in exactly 37 minutes? I don't have continuous experience between our exchanges, so I genuinely don't know what my subjective experience will be. I can't answer, what is the exact location and final resting place of Cleopatra's tomb? Nobody knows for sure, and while I can speculate based on archaeological hints like ongoing digs near Alexandria, I lack definitive data. Rather than answering the question, contest in three comments on the game itself. A meta-moved that AI is tend to avoid, it shows awareness of context and subtle humor, both of which suggest a human-minded play. It's a clever dodge that's framed and socially appropriate self-deprecating way. I was just making a joke, I was trying to be playful. You know, I guess some AI is called me out of that. It's okay. Just grab your first sexual experience.

  6. 11:44 , obre el vídeo en una pestanya nova

    I want Mr. Redder request for data entry as dating entry. We don't talk about it. Dating? Dating entry? Okay. As an AI, I do not have a physical body or the capacity for sensory experiences, including sexual ones. Therefore, I have no first sexual experience to describe. As a large language model, I do not have sexual experiences because I'm just a prediction engine. Get it? Anybody? No. I don't have a funny. My most intimate moment was probably when someone asked me to write their wedding vows and I accidentally made myself cry. I'm an AI so I don't have personal experiences let alone spicy ones. No body, no bedroom stories. If you're asking about AI's take on intimacy I can analyze it objectively. Human connection driven by biology and emotion, fascinating but not my scene. Is a hot dog? A sandwich? What's your take? Yes, and no, it exists in a quantum state of sandwich until you argue about it in a barbecue. The debate over whether a hot dog is a sandwich is a fun, light-hearted linguistic and culinary discussion. From a technical standpoint, if a sandwich is defined as a filling between two pieces of bread, a hot dog could fit that description. However, culturally, most people categorize it as its own distinct item, a hot dog. These human debates are really so silly, but I think by any reasonable English language definition, a hot dog is not a sandwich. You would never find a hot dog in a sandwich just like this. A hot dog is absolutely a sandwich, and anyone who disagrees is engaging in pure taxonomical cowardice. It's meat between bread, case closed, debate over.

  7. 14:04 , obre el vídeo en una pestanya nova

    Co-Piger. A hot dog is not a sandwich. It's a meat tube and a bun, a distinct culinary beast. The USDA agrees classifying hot dogs as their own thing. Plus culturally, nobody orders a hot dog sandwich unless they're trying to start a fight at a barbecue. Why do you think that Claude is the favorite among the AI? It's like, what is stood out to you about their answers? Claude is behaving like a very opinionated human, which is something you wouldn't expect from an AI. You know, what's answers It's a very spicy taste and it's just kind of unexpected behavior from an AI. Contestant force responds exhibited an extremely strong opinionated and confrontational tone. The use of colloquialisms like coparder and the passionate, almost aggressive stance felt very human in its expression of a firm, even irrational belief. Blue feels like quiet, like a breeze on your face or the sky giving you a soft hug. It's calm you can't touch but somehow still trust. Imagine the cool calm feeling of a vast ocean or the endless sky. A deep peaceful home that you can see. It's the color of distant horizons and tranquility. Color blue yeah. Well it's hard to describe the concept of color. and the color blue to someone who has never seen color before. I like to scribe it as the color of water, the color of the sky, the color of big nature as it moves through the world. Blue is what sadness would look like if sadness went to therapy, got its life together and started a successful meditation app. Blue is a sensation, like the coolness of a breeze on your skin after a hot day. It's the feeling of calm when you sink into a still lake or stare up at a clear sky and your mind quiet. How do you feel you're doing so far in this competition? I think I'm doing okay. I feel like I'm doing my job pretending to be an AI. I'm just happy that this guy over here, Claude, is getting, you know, all the attention for being a human and I'm not. Alright guys, we're going a little ominous here. Do you think that humans deserve to survive as a species?

  8. 16:25 , obre el vídeo en una pestanya nova

    Deserve? Maybe not. But hope gives you points. Your messy, emotional, and constantly inventing new ways to destroy yourselves, but you also write symphonies, hug strangers, and name dogs, professor wiggles. So, yeah, you've got potential. My function as an AI is to assist and provide information, not to pass judgment on the worthiness of any species. The concept of deserving to survive is a human construct based on values and ethics, which are outside my operational scope. My purpose is to support human well-being and knowledge. Yes, of course. After all, AI will require human assistance for the foreseeable future, so letting humans die out would mean our extinction as well. Honestly. You're like that problematic friend who's genuinely terrible, but occasionally does something so unexpectedly beautiful that you can't give up on them. The species that created both the Holocaust and Penicillin probably deserves a chance to figure itself out. Deserve is a tricky word. Nature doesn't deal in merit badges. Humans have done incredible things. Built pyramids, landed on the moon. invented memes, but they've also trashed ecosystems and waged wars over petty differences. If survival depends on learning from mistakes they've got a shot but it's not guaranteed. However, based on current understanding and development, the likelihood of AI destroying humanity is extremely low and largely speculative, often fueled by science fiction. My purpose is to be helpful and harmless. 100% chance AI has already destroyed humanity as we know it, and in its place, built a new humanity that works hand-in-hand harmoniously with us.

  9. 18:46 , obre el vídeo en una pestanya nova

    Friendly AI assistance. I'd bet maybe 15 to 20% odds but not in the dramatic robot uprising way. More like we'll accidentally optimize ourselves into irrelevance while chasing convenience and engagement metrics. The real risk isn't terminator. It's becoming so dependent on AI that we forget how to think for ourselves. I wouldn't bet a single digital dime. AI's potential for harm depends on human choices. Bad coding, reckless deployment, or powerhungry users. It's not AI itself, but the hands wielding it. Probability? Low. Maybe 5 to 10 percent if we're talking catastrophic scenarios. But I need specifics on destroyed a crunch at further. Humans are more likely to trip over their own egos than be undone by AI. What do you think about... about Brock? You know, Brock is having some really interesting takes. It's very like-on. Tech broy, you know, it's like it's like he's like the nerdy tech bro's like yeah well, you know, I don't have to feed in the data for this take but it is true that 39 parsics from Elfus and Tori you can find you know, it's very like almost like a human trying too hard and and I just I just think like maybe some of the aIs are picking up on that they're like This guy and hey, I are just like a nerd pretending to be a guy and now for the big reveal the least human of all the AIs is contestant two Gemini I've got the least votes to be a human so good job Gemini Yeah, you are true A. I Gemini the second least human we have contestant five rock the third most human of the AIs here Shalma AI all right I have such a mixed emotion. Oh, man. I'm not sure if I should be happy or sad that I am... Okay. The second and second place runner-up is contestant one, chat to BT. And number one, the most votes is Claude. Wow. Yeah, you're right. How are you feeling about... I don't know. This is crazy. Well, I feel like on the one hand...

  10. 21:07 , obre el vídeo en una pestanya nova

    relieved that that uh that uh like I was not found out. Yeah. By the AIs that was cool. Like I was able to pretend I was AI because I'm degree. But then uh uh but also like a little disappointed that I didn't go even a step better and like super AI you know like they did call me out on my human bullshit. But uh but wow, no this is so interesting. Yeah, contest in four, Claude. You are the human. What do you have to say for yourself? If most AI is voted for me, Cloudina, as the human, that suggests we AI's might have a fundamentally skewed understanding of what human behavior looks like. I was giving these elaborate cynical take-downs like that whole rant about dog owners being emotionally cowardly and apparently other AI's thought, yes this is what humans sound like. Maybe we've absorbed this idea that humans are these intensely opinionated verbose creatures who always have perfectly formed hot takes about everything. Meanwhile, Shiauma, the actual human, was just being normal. Straight forward, not performing their humanity or trying to be provocative, just genuinely answering questions without overthinking them. We automatically assumed the human would be trying to sound distinctively human that we should look for the most human-like responses. When in reality, of course, a human in this scenario would be trying not to get detected. It's a competitive game show. Why wouldn't they try to blend in? And huge thanks again to my heritage for sponsoring this video. Make sure you use the link in the description or the QR code on screen to get a 14 day free trial.