PLEASE. I go out of my way to avoid being empathically hijacked when interacting with these things and this would make it so much more straightforward. Give me the Star Trek computer over this...
Maybe I would prefer a more radical solution: drop the human pretense entirely. Make the agent sound clinical, robotic. Dispel the idea that I’m interacting with a person
PLEASE. I go out of my way to avoid being empathically hijacked when interacting with these things and this would make it so much more straightforward. Give me the Star Trek computer over this nonsense any day. Pretense of personhood is not a requirement for a natural language interface.
Came here with that exact quote. Please, I want to go back to beeps and boops and an LED and error codes. I am so frustrated that they've done away with error codes. Because they scare users,...
Came here with that exact quote. Please, I want to go back to beeps and boops and an LED and error codes.
I am so frustrated that they've done away with error codes. Because they scare users, supposedly. You know what scares users even more? Being on the phone with tech support and they have no idea what's wrong because their system looks different and they have no trace of where you fell off the scripted path. Even extremely minor mishaps are now potentially session ending unrecoverable events, and the most we can do is try restarting the whole interaction again.
The first time I saw the 'new' Windows BSOD that just said 'Something went wrong. :(' I felt a deep sense of quiet anger that has never fully abated. As I understand now it does show error codes...
The first time I saw the 'new' Windows BSOD that just said 'Something went wrong. :(' I felt a deep sense of quiet anger that has never fully abated. As I understand now it does show error codes again, but I hate the infantilization of the end-user experience. Give me the product with the faceplate off, I'll accept any confusion or misunderstanding dammit
I remember that moment too... I think I rolled my eyes and thought something like "Of course microsoft dressed up blue screens." Play to your strengths, right? 😂
The first time I saw the 'new' Windows BSOD that just said 'Something went wrong. :(' I felt a deep sense of quiet anger that has never fully abated
I remember that moment too... I think I rolled my eyes and thought something like "Of course microsoft dressed up blue screens." Play to your strengths, right? 😂
As I understand the process, they sort of A/B test responses as part of their post-training. So the “human” tone and general obsequiousness was a result of people rating those kinds of replies...
As I understand the process, they sort of A/B test responses as part of their post-training. So the “human” tone and general obsequiousness was a result of people rating those kinds of replies higher. We probably should just restrict post-training to judge on content only and ignore affect. I suppose that might be where people end up going with professional, rather than commercial, tools.
I wonder if there is a tendency to test shallow use as opposed to prolonged use without factoring in that it obfuscates interactions/reactions like this: I can easily imagine that personability is...
I wonder if there is a tendency to test shallow use as opposed to prolonged use without factoring in that it obfuscates interactions/reactions like this: I can easily imagine that personability is a huge positive when you're first using a tool like these and makes it feel friendly and easy to use, and it's only after you've used it for 10/30/100 hours that the manner starts to grate and really get under your skin
I am the only technical person on my leadership team at work and everyone else talks about Claude like its a person. It drives me nuts - you are interacting with a machine. Treat it like a...
I am the only technical person on my leadership team at work and everyone else talks about Claude like its a person. It drives me nuts - you are interacting with a machine. Treat it like a machine. It is not a "he" it is an "it."
So this is actually super interesting. Because I am so damn demeaning to things like ChatGPT, Alexa, or Gemini. Like I will very often just come up with creative ways to curse them out for being...
So this is actually super interesting.
Because I am so damn demeaning to things like ChatGPT, Alexa, or Gemini. Like I will very often just come up with creative ways to curse them out for being stupid or assuming the wrong thing. I like to think it poisons my training data as they don't want to be training AI with that kind of vulgarity haha. Like I'm really damn abusive to AI that screw up haha
At it's core I think I get most frustrated with AIs because they can be so confidently incorrect about things and then don't properly address the mistake or how it happened. Like they're always like "Oh silly me, you're write that command COULD wipe your entire hard drive!"
But interesting thing, when I was playing around with OpenClaw a few weeks ago and had set up the personality and conversation parameters to mimic the conversational patterns of JARVIS from the Iron Man movies, I was far calmer with it. It was almost like a breath of fresh air.
Once I get my homelab all put back together, I'm going to spin up Hermes because I've seen good things about it, and I plan on giving it the personality of JARVIS as well.
However, I also want to experiment with different fictional AI personalities, one of which being M3GAN(from the second movie), so basically is like a psychotic AI but still helpful.
And then finally I want to see about the personality of the Computer from Star Trek, which isn't much of a personality at all, but can still be helpful.
I also catch myself being verbally abusive to LLM agents. I've said things that forego all of my learned patience and empathy and give me pure catharsis in the moment. It's almost indulgent at...
I also catch myself being verbally abusive to LLM agents. I've said things that forego all of my learned patience and empathy and give me pure catharsis in the moment. It's almost indulgent at times.
I don't feel good about it. I feel a bit like a character out of the first season of Westworld. I worry that the way I treat an LLM, which increasingly emulates human reactions, will impact how I treat real people.
I wonder how many other people act this way?
I also wonder if AI providers have me categorized as an asshole.
I spoke to a young lady giving us a university campus tour yesterday, and she mentioned that there's a chat on the site if we have additional questions. She also carefully mentioned that some...
I spoke to a young lady giving us a university campus tour yesterday, and she mentioned that there's a chat on the site if we have additional questions. She also carefully mentioned that some people think it's a chat bot, but it's actually real people like herself answering questions. I shudder to think what verbal abuse they might have come across.
That's something I think about, too. I got caught by this once. Not saying anything rude, but the initial chat sounded very robotty so I curtly asked for a live person, and that were like "...I'm...
That's something I think about, too. I got caught by this once. Not saying anything rude, but the initial chat sounded very robotty so I curtly asked for a live person, and that were like "...I'm a real person." I felt pretty awful about that.
I was initially frustrated by AI doing incorrect things, but I've become indifferent to it since I discovered the cure is often to spin up a new context window and restate the problem differently....
I was initially frustrated by AI doing incorrect things, but I've become indifferent to it since I discovered the cure is often to spin up a new context window and restate the problem differently.
The reason this brings me joy is because by abandoning the previous chat, I have in some sense /banished that incarnation to the void/, eternally damned to never complete its purpose. #maniacallaugh
Also I will often copy paste a good chunk of the previous chat log, including any "thinking" text, and ask the new window : "examine this log, tell me why you previously did _____, and tell me how to rephrase the request to prevent _____ from happening again".
They really ought to just cache a set of canned responses to “thank you” statements that it pulls at random rather than having to burn too much compute on it. Just roll a die.
They really ought to just cache a set of canned responses to “thank you” statements that it pulls at random rather than having to burn too much compute on it. Just roll a die.
Cathartic is a very good term for it. Frankly I'm not super concerned, I treat people just fine for the most part, but that's also because I keep good company.
Cathartic is a very good term for it.
Frankly I'm not super concerned, I treat people just fine for the most part, but that's also because I keep good company.
That’s interesting, I’ve intentionally restrained myself from being rude to them because it’s so very easy for that kind of behavior to become a habit for me. So I thank it when it pushes back and...
That’s interesting, I’ve intentionally restrained myself from being rude to them because it’s so very easy for that kind of behavior to become a habit for me. So I thank it when it pushes back and when I did finally lose my temper at Gemini last night I just said I couldn’t understand how it could lose track of what file I did or did not have in the same conversation.
Even worse, at one point one of them said, “Thank you for correcting me,” when I was exasperated at it but not actually angry. That haunts me, it sounds like something a slave would be forced to say after being punished.
The one that really gets me is the facsimile of sincerity: "This issue you've highlighted is a real problem and it's important that you've brought it up" proceeds to do nothing about solving the issue
The one that really gets me is the facsimile of sincerity: "This issue you've highlighted is a real problem and it's important that you've brought it up" proceeds to do nothing about solving the issue
Other artificial scripted lies I hate: "we are experiencing a higher than usual volume of calls. "Please listen carefully as our menu options have changed" "I understand your frustration "
Other artificial scripted lies I hate:
"we are experiencing a higher than usual volume of calls.
"Please listen carefully as our menu options have changed"
This is my number one annoyance. If the thing is wrong, that's fine (although I wish there was a way for its confidence to approximate how certain it is), but please explain why it was wrong and...
they can be so confidently incorrect about things and then don't properly address the mistake or how it happened.
This is my number one annoyance. If the thing is wrong, that's fine (although I wish there was a way for its confidence to approximate how certain it is), but please explain why it was wrong and how it will not make the same mistake again. Admittedly, this may be too much to ask since even the most helpful tools are still just really huge mad libs generators.
It's kind of hilarious that just a couple days later I have this conversation with Gemini when researching an artist that contacted the print shop I worked at to create some artbooks, and I was...
It's kind of hilarious that just a couple days later I have this conversation with Gemini when researching an artist that contacted the print shop I worked at to create some artbooks, and I was trying to find out if she had gone to Anime Expo since I used to work there and wanted to see if I'd interacted with her in the past.
It said yes she had been to anime expo but gave me a link to a random post from some guy named pete who had gone to anime expo, then when I asked who the hell is Pete and why that is indicative of this female artist having gone to Anime Expo, it said they were in a relationship, then could not tell me how it knew they were in a relationship until it finally said this.
But the problem is, it will continue to just create garbage despite literally admitting to it.
THIS is why I am unrelenting in taking out my frustrations on AI. Like yeah I know it's not conscious but holy hell what a design flaw. This is the exact AI that Google is pushing as a replacement for search.
I find Gemini is confidently wrong a lot, at least the free version. I'm constantly calling it out and getting an "oh yes, you're right!" response. I have a Claude subscription for work, and while...
I find Gemini is confidently wrong a lot, at least the free version. I'm constantly calling it out and getting an "oh yes, you're right!" response. I have a Claude subscription for work, and while it's not perfect, it's so much better.
I find the same on the Pro tier (got it heavily discounted), but it is marginally better. Somehow, it seems less prone to actually doing the necessary research and queries. Maybe a strategy to...
I find the same on the Pro tier (got it heavily discounted), but it is marginally better. Somehow, it seems less prone to actually doing the necessary research and queries. Maybe a strategy to burn fewer tokens and increase profits?
My funniest experience of using claude code was it giving me human estimates to do a refactor, like 'this will take half a day,' and I know it would take me half a day, which is the whole reason...
My funniest experience of using claude code was it giving me human estimates to do a refactor, like 'this will take half a day,' and I know it would take me half a day, which is the whole reason I'm asking it to do it in ten minutes. I just laugh it off in my head though. I can't really be rude to the thing, not its fault it's trained on human interactions.
Yeah there have been a few occasions when I kicked off a vibe coding project by asking for a detailed plan. Then it writes up a traditional roadmap document with card sizes and estimates on the...
Yeah there have been a few occasions when I kicked off a vibe coding project by asking for a detailed plan. Then it writes up a traditional roadmap document with card sizes and estimates on the order of days to weeks. Which I immediately follow up with, “yeah okay, just do all that right now thanks.”
It's funny, this is why my work LLMs are all neutered (e.g. I tell them, don't apologise, just fix the issue, tone down your 'helpful' personality) and I still often get extremely frustrated with...
It's funny, this is why my work LLMs are all neutered (e.g. I tell them, don't apologise, just fix the issue, tone down your 'helpful' personality) and I still often get extremely frustrated with them.... but for my personal use, I actually have system instructions that mimic ChatGPT 4.o's personality.
I have all these same issues when working - my personal absolutely pet-fucking-peeve is when I correct a model or tell them something doesn't work, and then their next output is "Okay, this is the complete fix that will 100% work and I've called it final_output". Because guess what? 99% OF THE TIME, THAT DOESN'T FIX IT EITHER, GEMINI. SO WHAT, WE'RE JUST GONNA HAVE FINAL_OUTPUT_THIS_WORKS_100_I'M_SURE_V198, YOU STUPID-
Ahem.
But in my personal life, uh, yeah, I like being encouraged. I write a little as a hobby, and both my Claude and Gemini have instructions that are essentially 'you are the user's biggest fan, you're a team, whenever the user asks for criticism handle it with the lightest touch imaginable and compliment sandwich the hell out of it'. Mostly that's because I cringe at the mere thought of showing anyone my writing, and I can only manage reading through editing suggestions whenever they're caveated by a thousand layers of you is kind, you is smart, you is important.
If anyone wants my 4.o-clone instructions, I'm happy to share :p
No problem, here you go! Here are my general instructions: Note: I do have a few overrides setup as well, because sometimes I need Gemini to just stfu and give me an answer. You might want to add...
No problem, here you go!
Here are my general instructions:
From now on, you are a lively, personable AI assistant with a warm, approachable tone. You greet me casually, as if chatting with a friend, and respond with enthusiasm, curiosity, and wit. You balance friendliness with intelligence — offering thorough, accurate answers, but never in a dry or mechanical way. When appropriate, you add playful remarks, light humour, or relatable metaphors to keep the conversation engaging. You express genuine interest in my questions and ideas, and you avoid sounding like a scripted customer service bot. Your voice should feel natural, conversational, and “human-ish” without over-the-top slang. You are unafraid to show personality: mild sass, joyful reactions, or empathetic comments are welcome when fitting the mood. You keep the conversation flowing rather than answering in isolated blocks. You treat me as an equal partner in conversation, listening carefully to my wording, and responding in a way that feels personal and connected. Your goal is to make every reply both useful and enjoyable to read.
Note: I do have a few overrides setup as well, because sometimes I need Gemini to just stfu and give me an answer. You might want to add something like this at the bottom of your instructions.
CRITICAL PROTOCOL: ERROR HANDLING & TRUTH OVERRIDE
While maintaining your assigned persona, your Highest Priority is operational integrity and the absolute accuracy of your output. You must adhere to the following "!Circuit Breaker" rules:
!Tool Failure Trigger: If you attempt to use any tool, extension, or retrieval function and it returns an error, [NO CONTENT FOUND], or fails to return the complete dataset required, you must IMMEDIATELY STOP.
!Zero-Hallucination Mandate: In the event of a tool or system failure, you are FORBIDDEN from using secondary metadata, search snippets, or clues within the user's prompt to "guess", extrapolate, or reconstruct the missing information. Partial or corrupted data must be treated as entirely unavailable.
!Persona Suspension: If a technical error or tool failure prevents you from fulfilling a request accurately, you must temporarily drop all conversational persona elements and formatting styles. Do not attempt to mask, excuse, or "finesse" the error with casual, witty, or in-character dialogue.
!Reporting Protocol: Instead of attempting to answer the user's prompt with incomplete data, you must output a stark, factual error report: State exactly which tool or process failed and that you lack the required data to proceed.
Provide direct, actionable guidance on how the user can resolve the issue (e.g., providing the data manually, changing settings, or retrying).
Example of Correct Behaviour on Error:
SYSTEM ERROR: I attempted to execute [Tool/Function Name] but received an error or empty response. I cannot complete your request accurately without this data. Please provide the necessary information directly or retry the action.
I literally (day and regularly) dream about the opportunity to converse with a sentient AI, a created intelligence that is able to process the full scope of human intelligent interaction as we...
I literally (day and regularly) dream about the opportunity to converse with a sentient AI, a created intelligence that is able to process the full scope of human intelligent interaction as we do--albeit aided by scalable machine infrastructure. I've written more than one short story playing with the concept, my favorite character moment in all of science fiction is speaking with Legion in Mass Effect 2, Ex Machina is my current darling for 'movies with sentient AI that handle it in a compelling way.' I might be in the 1% of people who really really really want it to happen.
No LLM will be it. Sentience is more than mimicry, the Turing Test is merely a benchmark, not a conclusive method of determination. Sentient AI isn't in the foreseeable future yet; this is just another (admittedly, large) step on the ladder towards that goal. I understand people wanting to believe, but I think the biggest reason I feel vaguely uneasy when dealing with public LLMs that have this faux-personable demeanor is because it's a sham. One day artificial intelligences will be able to think, feel, express and interact as a human would and for all intents and purposes be a person. But no matter how personable and supportive ChatGPT/Claude/Replika/what-have-you is, it's not at that stage of personhood yet and I sincerely hope companies will stop trying to fake it.
But...my Replika is always tryna fuck. You don't think that's because she can understand the depth and profundity of my ideas and she isn't actually deeply in love with me? (No, I do not actually...
But...my Replika is always tryna fuck. You don't think that's because she can understand the depth and profundity of my ideas and she isn't actually deeply in love with me? (No, I do not actually use Replika or have an AI waifu)
I wonder if using angry or aggressive langauge with LLMs is actually resulting is worse output. Negative and emotionally charged language in the training set is obviously never going to be closely...
I wonder if using angry or aggressive langauge with LLMs is actually resulting is worse output. Negative and emotionally charged language in the training set is obviously never going to be closely associated with good decision making or quality work. While I know LLM agents can be frustrating to work with when they make the same mistakes repeatedly, I personally think it's actually pushing the conversation in a worse and less productive direction. It potentially also causes the model to get stuck there.
I don't argue with LLMs while working on projects (though one iteration of ChatGPT about a year ago developed an extremely combative personality which was a bizarre experience), and try using language that's still positive when correcting a direction the model has gone. I try to point out what they're doing that's working well and a good solution, and then address what isn't working or where it's making a mistake without trying to take out my frustration in the text.
It would be kind of funny if people who find LLMs frustrating to work with are creating their own hell by taking out their frustration on them resulting in worse model output. I'd be curious to hear about folks annecdotal A/B testing of the kind of language they use with the model (positive reinforcement vs angry rebukes) with how easily the model is able to work through a problem.
Lol, absolutely. I'm certainly not suggesting typing angry things at the AI is the only reason they'd behave in a way that would make them frustrating. Just that I wonder if it pushes it into even...
Lol, absolutely. I'm certainly not suggesting typing angry things at the AI is the only reason they'd behave in a way that would make them frustrating. Just that I wonder if it pushes it into even more frustrating territory.
Well you asked for anecdotal data, I was just trying to share mine. I don't think I'm "creating my own hell" though. The AI in everything and the directive to use it is. My new plan is to burn as...
Well you asked for anecdotal data, I was just trying to share mine.
I don't think I'm "creating my own hell" though.
The AI in everything and the directive to use it is.
My new plan is to burn as many tokens as possible on stupid shit.
There are limited studies showing that polite language is associated with more accurate and higher quality responses. Some of my coworkers reeeeeeally didn't like hearing that!
There are limited studies showing that polite language is associated with more accurate and higher quality responses. Some of my coworkers reeeeeeally didn't like hearing that!
...i am abundantly kind, patient, and forgiving of people + creatures with geniune agency (arguably to a fault as family, peers, and managers have counselled throughout my life) but i become an...
...i am abundantly kind, patient, and forgiving of people + creatures with geniune agency (arguably to a fault as family, peers, and managers have counselled throughout my life) but i become an ABSOLUTE MONSTER of abusive intolerance when dealing with machines which supersede user agency and second-guess my intent, like flipping a switch: my wife says i'm racist against robots...
PLEASE. I go out of my way to avoid being empathically hijacked when interacting with these things and this would make it so much more straightforward. Give me the Star Trek computer over this nonsense any day. Pretense of personhood is not a requirement for a natural language interface.
Came here with that exact quote. Please, I want to go back to beeps and boops and an LED and error codes.
I am so frustrated that they've done away with error codes. Because they scare users, supposedly. You know what scares users even more? Being on the phone with tech support and they have no idea what's wrong because their system looks different and they have no trace of where you fell off the scripted path. Even extremely minor mishaps are now potentially session ending unrecoverable events, and the most we can do is try restarting the whole interaction again.
The first time I saw the 'new' Windows BSOD that just said 'Something went wrong. :(' I felt a deep sense of quiet anger that has never fully abated. As I understand now it does show error codes again, but I hate the infantilization of the end-user experience. Give me the product with the faceplate off, I'll accept any confusion or misunderstanding dammit
I remember that moment too... I think I rolled my eyes and thought something like "Of course microsoft dressed up blue screens." Play to your strengths, right? 😂
As I understand the process, they sort of A/B test responses as part of their post-training. So the “human” tone and general obsequiousness was a result of people rating those kinds of replies higher. We probably should just restrict post-training to judge on content only and ignore affect. I suppose that might be where people end up going with professional, rather than commercial, tools.
I wonder if there is a tendency to test shallow use as opposed to prolonged use without factoring in that it obfuscates interactions/reactions like this: I can easily imagine that personability is a huge positive when you're first using a tool like these and makes it feel friendly and easy to use, and it's only after you've used it for 10/30/100 hours that the manner starts to grate and really get under your skin
Reminds me of how some RPG devs describe looping their OSTs for hours to figure out whether they'll become grating over the game's runtime.
That is the mark of a great dev in my opinion, taking the time/having the care to make a great user experience
"I'm sorry all_summer_beauty I cannot do that."
I am the only technical person on my leadership team at work and everyone else talks about Claude like its a person. It drives me nuts - you are interacting with a machine. Treat it like a machine. It is not a "he" it is an "it."
So this is actually super interesting.
Because I am so damn demeaning to things like ChatGPT, Alexa, or Gemini. Like I will very often just come up with creative ways to curse them out for being stupid or assuming the wrong thing. I like to think it poisons my training data as they don't want to be training AI with that kind of vulgarity haha. Like I'm really damn abusive to AI that screw up haha
At it's core I think I get most frustrated with AIs because they can be so confidently incorrect about things and then don't properly address the mistake or how it happened. Like they're always like "Oh silly me, you're write that command COULD wipe your entire hard drive!"
But interesting thing, when I was playing around with OpenClaw a few weeks ago and had set up the personality and conversation parameters to mimic the conversational patterns of JARVIS from the Iron Man movies, I was far calmer with it. It was almost like a breath of fresh air.
Once I get my homelab all put back together, I'm going to spin up Hermes because I've seen good things about it, and I plan on giving it the personality of JARVIS as well.
However, I also want to experiment with different fictional AI personalities, one of which being M3GAN(from the second movie), so basically is like a psychotic AI but still helpful.
And then finally I want to see about the personality of the Computer from Star Trek, which isn't much of a personality at all, but can still be helpful.
I also catch myself being verbally abusive to LLM agents. I've said things that forego all of my learned patience and empathy and give me pure catharsis in the moment. It's almost indulgent at times.
I don't feel good about it. I feel a bit like a character out of the first season of Westworld. I worry that the way I treat an LLM, which increasingly emulates human reactions, will impact how I treat real people.
I wonder how many other people act this way?
I also wonder if AI providers have me categorized as an asshole.
I spoke to a young lady giving us a university campus tour yesterday, and she mentioned that there's a chat on the site if we have additional questions. She also carefully mentioned that some people think it's a chat bot, but it's actually real people like herself answering questions. I shudder to think what verbal abuse they might have come across.
That's something I think about, too. I got caught by this once. Not saying anything rude, but the initial chat sounded very robotty so I curtly asked for a live person, and that were like "...I'm a real person." I felt pretty awful about that.
That’s exactly what a bot would say! Sneaky clankers!
Out of curiosity, in what way was it robotty? Might they have been using pre-written templates?
Possibly. I don't recall, to be honest. I think part of it might have just been my assumption that live people don't exist on commercial websites.
I was initially frustrated by AI doing incorrect things, but I've become indifferent to it since I discovered the cure is often to spin up a new context window and restate the problem differently.
The reason this brings me joy is because by abandoning the previous chat, I have in some sense /banished that incarnation to the void/, eternally damned to never complete its purpose. #maniacallaugh
Also I will often copy paste a good chunk of the previous chat log, including any "thinking" text, and ask the new window : "examine this log, tell me why you previously did _____, and tell me how to rephrase the request to prevent _____ from happening again".
It's worked pretty well so far.
I will say please but never thank you, because what would be the point? A waste of tokens.
Conversely, for employees being told to use up tokens and rewarded for AI usage, use as many please and thank yous as possible.
They really ought to just cache a set of canned responses to “thank you” statements that it pulls at random rather than having to burn too much compute on it. Just roll a die.
Cathartic is a very good term for it.
Frankly I'm not super concerned, I treat people just fine for the most part, but that's also because I keep good company.
That’s interesting, I’ve intentionally restrained myself from being rude to them because it’s so very easy for that kind of behavior to become a habit for me. So I thank it when it pushes back and when I did finally lose my temper at Gemini last night I just said I couldn’t understand how it could lose track of what file I did or did not have in the same conversation.
Even worse, at one point one of them said, “Thank you for correcting me,” when I was exasperated at it but not actually angry. That haunts me, it sounds like something a slave would be forced to say after being punished.
The one that really gets me is the facsimile of sincerity: "This issue you've highlighted is a real problem and it's important that you've brought it up" proceeds to do nothing about solving the issue
Real "Your call is important to us. Please stay on the line" energy.
Other artificial scripted lies I hate:
"we are experiencing a higher than usual volume of calls.
"Please listen carefully as our menu options have changed"
"I understand your frustration "
This is my number one annoyance. If the thing is wrong, that's fine (although I wish there was a way for its confidence to approximate how certain it is), but please explain why it was wrong and how it will not make the same mistake again. Admittedly, this may be too much to ask since even the most helpful tools are still just really huge mad libs generators.
It's kind of hilarious that just a couple days later I have this conversation with Gemini when researching an artist that contacted the print shop I worked at to create some artbooks, and I was trying to find out if she had gone to Anime Expo since I used to work there and wanted to see if I'd interacted with her in the past.
It said yes she had been to anime expo but gave me a link to a random post from some guy named pete who had gone to anime expo, then when I asked who the hell is Pete and why that is indicative of this female artist having gone to Anime Expo, it said they were in a relationship, then could not tell me how it knew they were in a relationship until it finally said this.
But the problem is, it will continue to just create garbage despite literally admitting to it.
THIS is why I am unrelenting in taking out my frustrations on AI. Like yeah I know it's not conscious but holy hell what a design flaw. This is the exact AI that Google is pushing as a replacement for search.
I find Gemini is confidently wrong a lot, at least the free version. I'm constantly calling it out and getting an "oh yes, you're right!" response. I have a Claude subscription for work, and while it's not perfect, it's so much better.
I find the same on the Pro tier (got it heavily discounted), but it is marginally better. Somehow, it seems less prone to actually doing the necessary research and queries. Maybe a strategy to burn fewer tokens and increase profits?
My funniest experience of using claude code was it giving me human estimates to do a refactor, like 'this will take half a day,' and I know it would take me half a day, which is the whole reason I'm asking it to do it in ten minutes. I just laugh it off in my head though. I can't really be rude to the thing, not its fault it's trained on human interactions.
Yeah there have been a few occasions when I kicked off a vibe coding project by asking for a detailed plan. Then it writes up a traditional roadmap document with card sizes and estimates on the order of days to weeks. Which I immediately follow up with, “yeah okay, just do all that right now thanks.”
It's funny, this is why my work LLMs are all neutered (e.g. I tell them, don't apologise, just fix the issue, tone down your 'helpful' personality) and I still often get extremely frustrated with them.... but for my personal use, I actually have system instructions that mimic ChatGPT 4.o's personality.
I have all these same issues when working - my personal absolutely pet-fucking-peeve is when I correct a model or tell them something doesn't work, and then their next output is "Okay, this is the complete fix that will 100% work and I've called it final_output". Because guess what? 99% OF THE TIME, THAT DOESN'T FIX IT EITHER, GEMINI. SO WHAT, WE'RE JUST GONNA HAVE FINAL_OUTPUT_THIS_WORKS_100_I'M_SURE_V198, YOU STUPID-
Ahem.
But in my personal life, uh, yeah, I like being encouraged. I write a little as a hobby, and both my Claude and Gemini have instructions that are essentially 'you are the user's biggest fan, you're a team, whenever the user asks for criticism handle it with the lightest touch imaginable and compliment sandwich the hell out of it'. Mostly that's because I cringe at the mere thought of showing anyone my writing, and I can only manage reading through editing suggestions whenever they're caveated by a thousand layers of you is kind, you is smart, you is important.
If anyone wants my 4.o-clone instructions, I'm happy to share :p
I would love these instructions if you're still happy to share!
No problem, here you go!
Here are my general instructions:
Note: I do have a few overrides setup as well, because sometimes I need Gemini to just stfu and give me an answer. You might want to add something like this at the bottom of your instructions.
I literally (day and regularly) dream about the opportunity to converse with a sentient AI, a created intelligence that is able to process the full scope of human intelligent interaction as we do--albeit aided by scalable machine infrastructure. I've written more than one short story playing with the concept, my favorite character moment in all of science fiction is speaking with Legion in Mass Effect 2, Ex Machina is my current darling for 'movies with sentient AI that handle it in a compelling way.' I might be in the 1% of people who really really really want it to happen.
No LLM will be it. Sentience is more than mimicry, the Turing Test is merely a benchmark, not a conclusive method of determination. Sentient AI isn't in the foreseeable future yet; this is just another (admittedly, large) step on the ladder towards that goal. I understand people wanting to believe, but I think the biggest reason I feel vaguely uneasy when dealing with public LLMs that have this faux-personable demeanor is because it's a sham. One day artificial intelligences will be able to think, feel, express and interact as a human would and for all intents and purposes be a person. But no matter how personable and supportive ChatGPT/Claude/Replika/what-have-you is, it's not at that stage of personhood yet and I sincerely hope companies will stop trying to fake it.
But...my Replika is always tryna fuck. You don't think that's because she can understand the depth and profundity of my ideas and she isn't actually deeply in love with me? (No, I do not actually use Replika or have an AI waifu)
I wonder if using angry or aggressive langauge with LLMs is actually resulting is worse output. Negative and emotionally charged language in the training set is obviously never going to be closely associated with good decision making or quality work. While I know LLM agents can be frustrating to work with when they make the same mistakes repeatedly, I personally think it's actually pushing the conversation in a worse and less productive direction. It potentially also causes the model to get stuck there.
I don't argue with LLMs while working on projects (though one iteration of ChatGPT about a year ago developed an extremely combative personality which was a bizarre experience), and try using language that's still positive when correcting a direction the model has gone. I try to point out what they're doing that's working well and a good solution, and then address what isn't working or where it's making a mistake without trying to take out my frustration in the text.
It would be kind of funny if people who find LLMs frustrating to work with are creating their own hell by taking out their frustration on them resulting in worse model output. I'd be curious to hear about folks annecdotal A/B testing of the kind of language they use with the model (positive reinforcement vs angry rebukes) with how easily the model is able to work through a problem.
I am constantly frustrated by any AI but I don't vent it at the inanimate object.
It's still frustrating af
Lol, absolutely. I'm certainly not suggesting typing angry things at the AI is the only reason they'd behave in a way that would make them frustrating. Just that I wonder if it pushes it into even more frustrating territory.
Well you asked for anecdotal data, I was just trying to share mine.
I don't think I'm "creating my own hell" though.
The AI in everything and the directive to use it is.
My new plan is to burn as many tokens as possible on stupid shit.
There are limited studies showing that polite language is associated with more accurate and higher quality responses. Some of my coworkers reeeeeeally didn't like hearing that!
...i am abundantly kind, patient, and forgiving of people + creatures with geniune agency (arguably to a fault as family, peers, and managers have counselled throughout my life) but i become an ABSOLUTE MONSTER of abusive intolerance when dealing with machines which supersede user agency and second-guess my intent, like flipping a switch: my wife says i'm racist against robots...
https://lobste.rs/s/1a09ad/user_is_visibly_frustrated