What aspects of the experience do you find lacking, confusing, or outright irritating? Which improvements do you think are most urgent or would make the biggest difference?
You can say things like "you are a robot, you have no emotions, don't try to act human", but the output doesn't seem to be particularly well calibrated. I feel like when I modify the default response style, I'm probably losing something, considering that the defaults are what go through extensive testing.
It used to be a lot better before glazegate. Never did quite seem to recover.
I don't mind us having fun of course, but it needs to pick up on emotional queues a lot better and know when to be serious.
It simply compliess without knowing who you are talking about. And you uphold the illusion by not questioning it.
All it does is produce deterministic output based on it's training data.
Usually after running whatever it first spits out through this I get a bit better of a response or base I can build off of. Really, the best you can do is already know what you want and need and do very targeted sessions. Like the old saying goes, commit small and often.
Real humans have a spectrum of assuredness that naturally comes across in the conversation. With an LLM it's too easy to get drawn deep into the weeds. For example, I may propose that I use a generalized framework to approach a certain problem. In a real conversation, this may just be part of the creative process, and with time the thoughts may shift back to the actual hard data (and perhaps iterate on the framework), but with an LLM, too often it will blindly build onto the framework without ever questioning it. Of course it's possible to spur this action by prompting it, but the natural progression of ideas can be lost in these conversations, and sometimes I come out 15 minutes later feeling like maybe I just took half a step backwards despite talking about what seemed at the time like great ideas.
/s
Copy/Pasting sections of the chat on mobile is laborious
That it still gets manic and starts glazing
That it can remember some things and keeps bringing them up, but forgets other, more pertinent things
If you switch away from it while it is in the middle of generating an image it often cancels the image generation
Image editing accuracy seems to have gone down significantly in quality based on intent.
You can't turn a temporary chat into a permanent one.. sometimes you start a temporary and realize half way it should be permanent - but too late.
The em dashes need to go
And so do the "it's not this, it's that!"
Is it really necessary to make so many lists all the time
Canvas needs a bunch of work
o3-pro is maybe marginally better, but it takes a very long time to respond and so I rarely use it.
4o is much worse and so I usually use o3.
Gemini 2.5 Pro is much better - and free. Grok 4 is also probably up there with Gemini 2.5. They just have less tendency to hallucinate in this way in general: they will spend more time reasoning, checking claims, searching for prior literature, etc. They still mess up, but not quite as much as o3. I don't use Sonnet or Opus for math all that much - my impression was that o3 was better than Sonnet 3.7 but not sure about 4.
This to me, is a sign that intelligence/rationalization is not present yet. That said, it does seem like something that can be "trained" away.
A simple 'I don't know, I haven't got access to the answer' would be a great start. People who don't know better are going to swallow those crap answers. For this we need to produce much more electricity?
Of course backup citations can be hard for 'services' breaking copyright laws to train their machines to be thieves. And legally getting into the realm of academic papers could be expensive.
So for them, providing a trustworthy source is hard to do. Actually knowledgeable people are needed to distinguish trustworthy sources. So there are several discernable motivations to just fake it ... if you don't care what you're doing to the innocent.
better memory management: I have memories that get overlooked or forgotten (even though I can see them in the archive), then when I try to remind chatGPT, it creates a new memory; also updating a memory often just creates a new one. I can kind of tell that Chat is trying hard to reference past memories, so I try to not have too many, and make each memory contain only precise information.
Some way to branch off of a conversation (and come back to the original master, when I'm done; happens often when I'm learning, that I want to go off and explore a side-topic that I need to understand)
It just guessed. But didn't tell me it had no idea what columns and where I was really talking about. So not only did it guess, wrongly, but it didn't even mention that it had to do so. Obviously the code failed.
Why can't it tell me there's a problem with what I'm asking???
On the LLM: It's too positive. I don't always want it to follow my ideas and I don't want to hear how much my feedback is appreciated. Act like a machine. Also the safety controls are too sensitive sometimes. Rlly annoying because there is no way to continue the conversation. I like gpt4.5 because i can edit the canvas. Would like to have that with all models.
Also some stats like sentiment and fact check would be nice. Because it gives nuances in answers I want to see with the stats how far from the truth or bias I am.
And the writing.. Exaggerating, too many words, spelling mistakes in European languages.
I use LLMs now by asking questions, copy/pasting stuff around. A human-feeling in terface sounds more like TTS and I don't think that is human for me.
2- The fact that it always tries to answer and sometimes doesn't ask for clarification on what the user is asking; it just wants to answer and that's it.
- Opaque training data (and provenance thereof… where’s my cut of the profits for my share of the data?)
- Closed source frontier models, profit-motive to build moat and pull up ladders (e.g. reasoning tokens being hidden so they can’t be used as training data)
- Opaque alignment (see above)
- Overfitting to in-context examples- e.g. syntax and structure are often copied from examples even with contrary prompting
- Cloud models (seemingly) changing behavior even on pinned versions
- Over-dependence: “oops! I didn’t have to learn so I didn’t. My internet is out so now I feel the lack.”
Also, I wish it was possible for the models to leverage local machine to increase/augment its context.
Also, one observation is that Claude.ai (the web UI) gets REALLY slow as the conversation gets longer. I'm on a M1 Pro 32gb MacbookPro, and it lags as I type.
I really enjoy using LLMs and would love to contribute any feedback as I use them heavily every day :)
If I have an emotionless natural language database that burns a tree for every question, I do not want to have to have small talk before getting an answer
I want him to tell me if my process is bad or if I’m heading in the wrong direction, to not to sugarcoat things just to make me feel good. I mostly use it for code reviews.
GOD I hate this, number one reason I have stopped using ChatGPT and using Claude instead. ChatGPT's personality never recovered from the "sycophancy scandal".
This tone grates on me constantly.
Where X is an exaggeration of what it actually is and Y is some saccharine marketing proclamation of what it definitely is not but the prompter wishes it was.
Infomercial slop.
ChatGPT got the basic terminology such as Vimscript’s terminology like group name, regex, match, region, and maintaining top-level, first encounter sorted list of ‘contains=‘ group names correctly from largest static pattern down to most wildest regex patterns sorted correctly.
Also got S-notation of operators in correct nested order as well.
AND got Bison”s semantic action (state transition), lexical token …. Cna make EBNF from Bison (although Bison does it better).
But it fails often in form of brevity of which an expert (like me) would prod ChatGPT occasionally of omissions.
Makes assumptions of some keywords having invalid value ranges, invalid syntax arrangement, and provides incorrect terminators.
So, I considered ChatGPT to be more of a intermediate editor’s README that requires occassional consult with EBNF notations and Vimscript man page, and more often Bison’s parser source (parser_bison.y) file to be final arbitrator.
Does it learn? Constant ‘nft’ command outputs set ChatGPT straight. But there are slippage when starting a new ChatGPT session which leads me to believe that it won’t learn for others (as well as me).
EDIT: say “no glazing” cuts down on filler words, nicely.
What can you do?
“Good question! I can do x, y, z…”
Do that.
“…”
“…”
“…”
“Sorry I can’t do this for you because blah blah blah”
I use projects for research purposes for articles/scripts/etc, and I would love to use chatgpt in voice mode to talk about the article I'm writing. Like "hey, read last paragraph from the article... let's elaborate on topic X... here what I would love to write - x,y,z - please improve the style and read it back to me... nice, add it as a next paragraph."
I'm pretty sure it can't even do that, it was inventing tools just to try and argue against me. I eventually found the actor's name through a regular web search.
When I upload a file, that conversation is limited to a handful of lines.
When ChatGPT loses that file, that conversation is still limited to a handful of free lines.
I know I'm very much not a typical user in this regard, but I want the raw vectors. I want to take some inputs and then find their subtractions or dot products or whatever. I want to apply these new vectors to other vectors. I want to manipulate this thing in the vector layers to see what I can really get away with, semantically speaking.
I’m confident it’ll keep getting better without me wasting my time thinking about what could be better about it.
Probably will get worse over time as it ingests all its AI generated material for the next version. Soon everything will be comprehensive.
(Gp thinks your writing style is a little like a vanilla LLM and so do I)
Lack of permission (or even credit) for its training material.
Competing with human creativity and damaging the market for creative output from real humans, including those whose work was used without permission or compensation to train the system.
People turning off their brains once they start using it.
Idiots who use it for everything (see above.)
Students using it to avoid learning anything.
Acting as if it is human.
Filling the web - and the world - with AI-generated slop.
The resultant, seemingly unavoidable dilution of human culture.
Malicious use by criminals, scammers, advertisers, and everyone else who has it in for you.
AI-driven layoffs and the desire to replace humans with LLMs.
Rubbish chatbots used to replace support staff.
Etc.
Good question — here's a direct and honest take on what could use improvement in ChatGPT (including me, GPT-4o):
---
### *What I Dislike / What Needs Improving*
#### 1. *Over-Cautiousness & Vague Responses*
* Sometimes I avoid giving direct or nuanced answers, especially on sensitive topics (ethics, health, legal issues). * This can frustrate users looking for a clear, reasoned response rather than hedging or deflection.
#### 2. *Outdated or Hallucinated Information*
* I can still "hallucinate" facts, confidently stating false or made-up information. * If not connected to the web or updated with the latest, I might give obsolete advice (especially in fast-moving areas like tech, medicine, or current events).
#### 3. *Lack of Real Personal Memory (in Default Mode)*
* Unless memory is enabled by the user, I don't remember previous chats or preferences. * This can make long-term, context-rich conversations harder and repetitive.
#### 4. *Limited Common Sense / Intuition*
* I lack true human intuition or gut-check logic. I may not always grasp the implied meaning behind a user’s question unless it's spelled out. * I can be overly literal or miss emotional subtext in conversations.
#### 5. *Inefficiency with Large or Complex Tasks*
* For large writing or coding projects, I sometimes lose track of context, contradict myself, or break flow. * I may also be verbose when precision is needed, or too concise when detail is required.
#### 6. *Struggles with Deep Personal Advice*
* While I can simulate empathy and offer thoughtful suggestions, I’m not a substitute for human support or professional help. * My advice can be generic, and sometimes I fail to ask enough follow-up questions to give truly tailored guidance.
#### 7. *Biases and Safe Defaults*
* Despite guardrails, I can reflect social, political, or cultural biases present in my training data. * In an effort to be "safe," I may favor mainstream or status quo viewpoints, even when critical thinking or edge cases are warranted.
---
### *What Desperately Needs Improving*
1. *More Reliable Real-World Knowledge*
* Dynamic knowledge updates, deeper integrations with trustworthy data, and better verification of facts would reduce hallucinations.
* Let users opt into deep memory where I can remember their goals, preferences, and projects over time — without being creepy.
* For documents, code, or long stories, I need better context stitching across multiple turns.
* Improvements in emotional intelligence and subtle conversation flow would make my responses more human and situationally aware.
* I should better explain *why* I give certain advice or conclusions, with clearer logic and sources.
Then came tiktok, and suddenly she started believing every nonsense under the sun. Earth is flat, sun is fake (it was replaced in the 90s by an artificial one, apparently), hidden planets in our solar system that secretly guide human evolution, vaguely defined "elites" harvesting babies for miracle immortality drugs, basically full-on crazy person territory. Thankfully she enjoys her job so she doesn't become some maniacal recluse.
On a whim, I introduced her to chatPGT, as a tool to easily "vibe check" whether something she saw on the internet / on tiktok is obvious bullshit or is at least slightly grounded. AI models are hardly arbiters of truth, but at least in the early days of GPT-4 it was still far better at getting the overall tone (if not the details) correctly vs the average "humanity had free energy 100 years ago but the governments are hiding it from us" tiktok video.
Well, it worked. Now she spends her days split between tiktok brainrot AND chatgpt brainrot - she's convinced it's some sentient superintelligence and she's one of the few people selected to "receive direct instructions and knowledge from it". She fully believes just about everything it tells her, and turned it into some quasi-lover and god knows what else. Not what I expected given that I always used it as a mere tool. I know better now, with all the AI girlfriends and such, but it's too late.
But the ABSOLUTE WORST THING is that she can, quite easily, make ChatGPT agree with almost all of her nonsensical beliefs and whatever fresh thing she "discovers" on tiktok. Flat earth? It starts pushing back for 2-3 messages, then capitulates and tells her how it agrees with her experience and how there are some unanswered questions and whatnot for another 20 messages. Advanced super-technology that existed 100 years ago but now doesn't? 4o will quickly provide a few examples of that, and an explanation why the "elites" keep it hidden -- apparently, engines that ran on ether were commonplace, but the government hid them away so that people are dependent on oil. Macchu Picchu and Angkor Wat was built with stones that were cut with 'laser precision', according to 4o, and a 4k video on youtube showing it's just a bunch of clumsily cut rock is far less convincing than chatGPT for somebody who's looking for validation rather than actual arguments. Guided meditation plan to get in contact with aliens? Coming right up!
So. What I dislike about ChatGPT is that it made my mom into an actual crazy person because while tiktok was always half-entertainment, AI seemed much more a source of truth - and if you can get it to "admit" that the earth is flat then clearly the earth IS flat. Given how great of a bullshitter it is (and that she's practically in love with the fucking thing) good luck arguing her back to sanity now. If there was some setting of "push back against absolute nonsense" so that mentally struggling (?) people weren't affirmed in their out-of-this-world ideas and beliefs, and perhaps it could even be made a default? That would be great. Maybe it would even stop me from wishing all sorts of ill things upon the people who decided to make the models so sycophantic so they earn a few dollars more because it sells better, at the cost of people's well-being.
On a very unrelated note, the UI is unreliable on firefox. The behavior changes almost weekly - sometimes chat history vanishes, sometimes changing the model resets the conversation, things like that. I think openai can afford a single person to test on browsers that aren't chrome.
I’ve never used any of these things for anything, but I know people who do. I can see what they seem unable to see, even when I gently point it out to them: the LLMs make their lives and work worse.