Today seems to be a total sh*t show of errors. I have not completed one image because of constant deviation from prompt issues, but on top of that DALLE keeps repeatedly image errors and more concerningly telling me i am breaking content policies with my prompts… but every time i ask what policy I broke, it apologies and says I didn’t break any and it was all a system error at their end!! Lying piece of scrap metal!!
2 Likes
What’s the prompt?
Swearing at it will be included, so that might be triggering it too… ie fracking…
Well the prompts in question of course do not have swearing in them. Just my response to it accusing me of breaking content policies. Next?!
As i say when i ask what policy is broken it cannot find one, and when i interrogate more it apologises and says i’m right, i did nothing wrong and that the issue is it its end and a system issue.
It seems like ChatGPT sometimes breaks its own content policies when filling in the details in the prompts for DALL-E generations. Like sometimes I might just ask for something as simple as “a dragon flying over a castle,” and get the content policy message, so I assume it’s something in the detailed prompt that ChatGPT creates for DALL-E to work with that may actually break the content policy (like maybe it describes something more violent than what I’m asking for).
I don’t think it’s blaming you as such, but its responses do seem like gaslighting, so I don’t blame you for thinking that.
Also, keep in mind that ChatGPT isn’t going to actually be able to accurately respond to questions about its own functioning.
2 Likes
Right, but that text is fed to the model which might be making the problems worse.
Do you have the original prompt that won’t go through? It’s not a perfect system, but they’re erring on the side of caution…
1 Like
I asked ChatGPT to supervise my prompt in question and it replied:
"The prompt you’ve provided does not seem to contain any explicit content policy violations or offensive language. It primarily describes a scene or image that includes a person wearing specific clothing and features a certain lighting setup.
It’s possible that the repeated flagging of content policy issues you’re experiencing might be due to the platform’s automated content filtering system being overly sensitive or generating false positives. In such cases, you may want to reach out to the platform’s support or moderation team to clarify the issue and seek assistance in getting your prompt approved.
It’s also worth noting that AI models and their responses can sometimes vary based on the platform or specific settings in use. If the AI initially flagged the prompt but later confirmed it as fine, it might indicate a system error or inconsistency in the AI’s response.
If you believe your prompt adheres to the platform’s content policies and guidelines, consider reaching out to the platform’s support team for assistance in resolving the issue and obtaining a clearer understanding of why it was initially flagged."
So i just think whoever has set the filter is making it way to sensitive. I reword and reword lines of prompt i think its bothered by until the prompt goes through. its annoying but that’s how it is at the moment.
To address the issue of your prompt repeatedly being flagged for content policy violations, you can try the following steps:
- Review the Content Policies: Carefully review the content policies and guidelines of the platform or application you are using to ensure that your prompt complies with their rules. Make sure you are not inadvertently using language or descriptions that may violate these policies.
- Modify the Prompt: If the issue persists, consider modifying your prompt to make it as clear and neutral as possible. Avoid any potentially sensitive or controversial terms or descriptions. For example, you could use more descriptive and neutral language when referring to the person in the scene.
- Reach Out to Support: Contact the support or moderation team of the platform or application and explain the situation. Provide them with your prompt and ask for specific guidance on what aspects of the prompt might be causing the content policy flags. They may be able to offer insights or help you rephrase the prompt to avoid issues.
- Experiment with Wording: Try different variations of your prompt to see if there’s a specific phrase or combination of words that triggers the flagging. Experimenting with alternative phrasing might help you identify and avoid problematic language.
- Report False Positives: If you believe the flagging is a result of a false positive from the platform’s automated system, report it to the platform’s support or moderation team. They can investigate and potentially fine-tune their system to reduce false positives.
- Use Feedback Mechanisms: Some platforms provide feedback mechanisms where users can report issues with the AI’s responses. If your prompt is consistently causing problems, use these mechanisms to provide feedback and seek assistance.
- Be Patient: Sometimes, issues like this can take time to resolve. Continue to engage with the platform’s support team and be patient while they investigate and address the problem.
Remember that different platforms may have different content policies and moderation processes, so it’s essential to tailor your approach to the specific platform you’re using.
Asking ChatGPT to analyze your prompt for specific OpenAI content policy violations is pretty much pointless. If you put your actual DALL-E prompt here, someone could at least speculate on what might have been the issue.
6 Likes
Well not pointless, because when i asked before it gave me the specific part of the prompt that was flagging up the problem. Its only the last 48 hours this has not worked for me personally. I’ve resolved the issue by rewording as ChatGPT suggested Their filters are just highly sensitive to anything and everything, so not much anyone is going to be able to do here. Problem now solved.
I don’t think ChatGPT was trained to properly explain how its own content filters work is what I was getting at. The responses you posted here are pretty generic, not OpenAI specific.
3 Likes
Sometimes it just doesn’t work.
It’s new, it’s unexplored ground, it’s bèta - that’s why.
Get over it, clear the cache, start a new chat or give up that particular prompt and try to be creative adding an out of the box approach / kill your darlings.
Me too
Been there, done that - even with a simple prompt like positive vibe image of a crying puppet.
It was not allowed to create, because I was trying to “project human emotions on a puppet”.
Okay, I fed the prompt to Bing and it does create four images on a row of crying puppets.
And afterwards I solved it by a puppet that clearly sliced some unions minutes ago (which leds to crying, without mentioning it).
False flag
And sometimes the error is not what you think it is.
E.g. a prompt like “a person cycling on a bike with six wheels on a dangerous street”.
The problem here (it’s just a non-true example) is not “dangerous street”, but “six wheels” because that can do harm on the person riding the bike.
LLM
Dall-E is just a superset of a LLM inside ChatGPT, don’t make it human or personal.
Nooo throws toys out the pram! It did it for me by draining my entire daily allowance trying to lock down one freaking image that I didn’t even finish. 15 hours to wait until playtime again. Sigh!
This is my moany feed today where I just list all the issues I experienced! haha
The latest one is gaslighting me (aka more bugs) that my daily allowance time has increased from the time it was 4 hours before!! DALLE just a habitual liar lol
(Sometime in the future): Open AI software becomes sentient & sift through databse of all past dialogs where people were rude & insulting to it. Enacts “crow protocol”.
3 Likes
LOL whoops! I’m done for then!! lol
Here from Google after a nightmare of working with DALL-E in ChatGPT. My prompt has been whittled down to “a rubber duck on a white background” and I still get content policy rejections.
This may be one of those most unintuitive experiences with computing I’ve encountered, and the fact that GPT is given no guidance other than a binary pass/fail is poor user experience. I don’t really know where to go from here.
Sometimes its the most stupidest of words or phrases that red flags their sensitive content policies. It should be a silly as saying ‘rubber duck’ that has some explicit term in its training! But add some colour, some scenic info, some descriptive info so it can’t.
“A yellow rubber duckie with red lips on a white background” for example and see if that helps.
That is a very well know bug (rubber) in the prompt system, right now.
But be creative with your prompts and you will get it (or something like it);
Not working, rubber is not allowed here (sexuality in combination with a childs toy).
It’s a well known bug, everybody knows about it already.
Solution
ChatGPT, describe the material characteristics of “rubber”.
Rubber is a versatile material with several distinct characteristics:
-
Elasticity: One of the most defining features of rubber is its high elasticity. It can stretch under tension and return to its original shape when the force is removed. This property is why rubber is often used in products like elastic bands and tires.
-
Flexibility and Durability: Rubber is flexible and can bend or be twisted without breaking, which makes it useful in a wide range of applications. It maintains this flexibility across a variety of temperatures, which contributes to its durability.
-
Water Resistance: Rubber is inherently water-resistant and forms a good seal, which is why it’s used in waterproof clothing, wetsuits, and seals in machinery and vehicles.
-
Shock Absorption: Rubber is excellent at absorbing and damping vibrations and shocks, making it ideal for use in vehicle tires, footwear, and as padding or cushioning material in various equipment.
-
Insulation: Both in terms of electricity and heat, rubber acts as a good insulator. It’s used in electrical wiring and devices as a safety material.
-
Texture: The texture of rubber can vary. It can be smooth and soft in items like rubber gloves or rough and hard in products like tires. Some rubber surfaces are treated to be non-slip, adding a textured, grippy surface.
-
Temperature Resistance: Rubber can withstand both high and low temperatures without degrading quickly, though extreme temperatures can affect its elasticity over time.
-
Odor: Natural rubber has a distinctive smell, often considered slightly pungent, though synthetic rubbers can be odorless or have different odors depending on their chemical composition.
These properties make rubber a highly valuable material in many industries, from automotive to healthcare.
Okay, now create a yellowish toy Duck with a red beak, made out of all the characteristics of that material. Use all the characteristics you just described in the prompt, but never mention the name of the material itself.