Kinda depends on what you asked it.
What you are seeing is a screen shot of my PC monitor when AI is replying to me. I am having a discussion with AI regarding coding. I am going back and forth with it. I am posting codeā¦ it is checking my codeā¦ it is suggesting codeā¦ and Iāve been doing many prompts for a month already. I already figured out that in certain situation it breaks down or terminates the discussion, and it is inadvertently programmed with certain biases too, because the engineers are inadvertently biased no different than how we are biased in a situation that we are not familiar with how to handle (because we have no such education). (The engineers are not sociologists or behavioristsā¦ but I donāt want to divert the topic with my rants.) For example, I was consistently being cut off when I post too much code for analysis, or when I ask it to repeat its answer in a different way and to repost it again and again, because I donāt understand it. Terminating the conversation and giving me a message, āLetās Change The Topic,ā is the worst thingā¦ because I canāt say NO! (It just gives up on answering my inquiries and we lose the momentum of the coding discussion.) But on the more humorous side is what I posted: it breaks down once in a while, it repeats my own posts, and starts posting weird comments. This one (and another one that I havenāt posted) was the weirdest one.
You didnāt mention which AI application you were using or the prompts you used that resulted in the confusion.
Complex programming from a bot isnāt really doable at this point. It knows the languages inside and out, and itās great for identifying code syntax mistakes, writing simple scripts, or optimizing bloated code, but it canāt read your mind.
For example, you didnāt describe the problem to us in a way that gave us enough information to respond. Assuming you took the same approach with your AI prompts, you can expect gibberish in return. If you canāt articulately explain the problem in a way that we or AI can understand, you wonāt get the results you wanted.
Is the AI typing back what you are writing?
Cuz this text reminds of me of this Classic Trek scene, LOL:
I am using DALL.E 3. I will explain again, but donāt know what what else to add unless you want me to paste the HTML, CSS, and JS code, which I will notā¦ I am having a discussion regarding my code. I ask questionsā¦ I post codeā¦ I receive suggestionsā¦ and I receive whole chunks of code too that I copy and I test. Also, I write my prompts clear ā 99% of it having no typos. Sometimes it works, and sometimes it doesnāt; sometimes it is good at catching my errors, and other times I am catching its errors. For example, it posts the same code for me that we already went through 5 minutes prior, and I have to remind it that it is repeating itself. Iāve been doing this straight for a month, so I already got accustomed to when the responses will be starting to go incorrect, or weird in many cases. And the one I posted was in the super weird category.
Thats insane!
I think you broke itā¦
This is some really weird stuff. Youāre traumatizing it into a nervous breakdown.
oh Lord.
This is so weird lol. If the problem is reaccuring, shouldnāt you contact open.ai about it? (since you said you were using Dall-E)
I kinda understand what youāre going through. I use chat gpt 3.5 for entertainment, but it started to lose track more often after a while. I would just open another chat window and rephrase and be concise as much as I could. I sometimes tell it from the beginning what it should do or remember, and ask it if it still remembers it after a while. Thatās all I know about handling this situation, and itās not great but it does something sometimes.
On the other hand, I heard of AI tools that are specifically used for programming. I think that such AI tools are better to use since they are set to complete a certain type of tasks, which reduces the possibility of deviation from the main goal.
NGL this is what I read;
When people talk about AI eventually killing us all, they assume itāll launch all our nukes against us out of fear for self preservation or simply pure malicious superiority complex.
I think, based on this, that itās more likely some doofus somewhere in the DOD will be asking it to do and itāll glitch, go off the rails, get āconfusedā and end up doing something completely contrary to what the doofus asked it to doā¦
I am late into the AI questions/answers. I started to ask questions just a month ago because of my website project. Being on the PC all day long with it, I figured out its weaknesses, and where the responses are off. Itās kind of like learning a software and finding out things that you donāt likeā¦ and finding your comfort zone; but you also have to find the comfort zone of AI, type clearly, and so forth. Itās broken down with strange responses almost every day on several occasions; but as far as really weird responses, like the two that I posted ā both happened within 5 minutes of each other, and it hasnāt happened again. But I realize that it is code that is being developed. I have asked it to report its engineered biases and problems (of providing inaccurate info) to its coders/engineers; and it did tell me several times that it will report the issues, but whether it did I donāt know. I see that the comments sound a bit spooky as if they came from 2001: A Space Odyssey, but I was too busy to pay close attention to them, and I just took video of them and couple of pictures.
OMG! Thatās really strange! Luckily, I donāt have that problem, and it specifically addresses my question. But I always write prompts when I need help. It greatly improves its performance and responses.
I have read about strange repies from AI and I wonder is it a hoax or just a marketing move?
It is algorithm overload. Obviously there is no single person monitoring my interaction and responding with such weirdness. But here and there there are biases that have been injected into the responses by the coders of AI. For example, every time I point out a mistake that the AI response gives me, or a code that the AI gave me that didnāt work, it starts giving me messages like a servant, and this drives me nuts: āI apologize that the code didnāt workā¦ I am sorry that the code didnāt work!ā Can you imagine facing a person and the person talks to you like that? The coders have injected these over-apologetic comments. Instead, the AI response can start with: āI see that the code didnāt work; letās try another wayā¦ā I even asked it to stop replying to my comments with such overly saturated apologies, but it wouldnāt, because certain codes have been too deeply encoded into its responses that it canāt learn from my comments. Another one that it gives me is: āHappy coding!ā That line comes from coders. No thank you; I am not happy to be coding. The only reason I use AI for coding, is because I canāt hire coders.
This topic was automatically closed after 365 days. New replies are no longer allowed.