AI is kinda cool

[Zasz]

24 minutes ago, AvoidTheCrown said:

When I go to the neighborhoods kebab place and I tell them to make it without olives and extra spicy I don't go around claiming that I did the dish myself. That is an accurate and objective analogy, not the twisted version of reality you've made up to try to be right.

 

But that is not how using AI to create a mod has worked in my experience. You are refering to an extreme scenario where the user is expecting the full final package without doing anything. Your analogy doesn't describe the process at all.

 

So maybe tune down the hostility.

[AvoidTheCrown]

6 minutes ago, SilentAegis said:

But that's not what OP has done.

 

Do you think I'm stupid or something?

 

 

 

 

The nerve you have to have to say an analogy can't be objective when it's a 1:1 representation of what has happened is next level delusion. Congrats 👏

[AvoidTheCrown]

5 minutes ago, Zasz said:

Your analogy doesn't describe the process at all.

 

Yes it does

 

 

7 minutes ago, Zasz said:

So maybe tune down the hostility.

 

Crazy thing is you use AI and expect not only to not be judged but the same respect and appreciation as an actual creator. Keep larping all you want, it's not going to take away the fact that the model is doing all the work.

[SilentAegis]

2 hours ago, AvoidTheCrown said:

 

Do you think I'm stupid or something?

 

 

 

 

The nerve you have to have to say an analogy can't be objective when it's a 1:1 representation of what has happened is next level delusion. Congrats 👏

Alright, I think this conversation is over as if it continues, this thread will just spiral. You have been entirely hostile thus far. No, I never said you were stupid. I don't know why you would take anything I've said in such a way, but that's not the intent. I'll respond one more time and then cease here since you don't seem interested in actually engaging. 

Your screenshots reinforce my point. Sure, the language is blunt, but they're testing output, reporting failures, and referring the AI back to earlier successful implementations. This is already materially different from your kebab customer example since OP did have to do testing themselves and give feedback themselves. ChatGPT made mistakes and it needed OP's feedback to improve and get to a finished result. This is still an iterative workflow. That's not a "1:1" representation to walking into a shop, ordering exactly what you want, and getting exactly what you want without fail. Of course, an analogy can be objectively better or worse depending on how accurately the relevant features of two situations are compared. They're still not objective measurements, but admittedly, what I said was a bit over-stated on my part.

But... in this case, yours fails as a 1:1 comparison. As we can see here, that's not what happened. OP had to work with it to get a usable result, they couldn't just say "make this mod" and it worked perfectly. So no, your analogy isn't objectively accurate anyway. It's actually rather off-base. If anything, this supports what I said earlier that AI isn't really a magic "make-mod" button. 

So I'm just going to bow out since you've just undermined your own stance here, you don't seem interested in a conversation, and you've been needlessly hostile and defensive. In fact, it's a little interesting to me that the only two I can see that are clearly against AI have been dismissive and judgemental at the minimum and hostile at the worst, while those of us more open have been perfectly civil.
 

 

EDIT: Also, just realized that you're wrong on it "not following a recipe", I think my example was fairly apt. A recipe is just externalized knowledge as instructions for producing a result. Which is closely related to what's been discussed throughout this thread. "Using knowledge someone else made", right? Obviously, my example isn't perfect, a recipe doesn't prepare food for you, whereas AI can perform part of the implementation itself, but it's certainly more accurate than yours.
 

Edited September 23 by SilentAegis

[xXDeliteXx]

1 hour ago, AvoidTheCrown said:

 

This is as much of a desperate attempt to try to be right as the other dude's bible you are quoting and it's simply a failed attempt to cope with mediocrity. None of that describes what is happening. It's not "following a recipe". When I go to the neighborhoods kebab place and I tell them to make it without olives and extra spicy I don't go around claiming that I did the dish myself. That is an accurate and objective analogy, not the twisted version of reality you've made up to try to be right.

Paintings don't paint themselves.

[AvoidTheCrown]

1 hour ago, SilentAegis said:

Your screenshots reinforce my point.

 

You are right. My screenshots clearly "undermine my own stance" because op tasted the kebab and said "you know what? Let's tune down the spice", thus being the one that made the kebab. You are clearly a desperate person in need to be right so right you are. I hope you get better and if not, you can always ask ChatGPT to help you with it.

[SilentAegis]

13 minutes ago, AvoidTheCrown said:

You are clearly a desperate person in need to be right so right you are.

...lol
 

 

3 hours ago, SilentAegis said:

I think their narrow point of transparency in utilizing AI in the creation of a mod is fair.

2 hours ago, SilentAegis said:

As I said in my first paragraph, you still have a point.

2 hours ago, SilentAegis said:

They're still not objective measurements, but admittedly, what I said was a bit over-stated on my part.

2 hours ago, SilentAegis said:

In fact, it's a little interesting to me that the only two I can see that are clearly against AI have been dismissive and judgemental at the minimum and hostile at the worst, while those of us more open have been perfectly civil.

[hogsmaws]

is she still bitter? sigh...

 

so i was using AI to edit a hair mesh and it walked me thru installing Blender and a ton of commands that i would not have learned easily. it really is a learning tool.

 

[hogsmaws]

3 hours ago, AvoidTheCrown said:

 

Do you think I'm stupid or something?

 

 

 

 

The nerve you have to have to say an analogy can't be objective when it's a 1:1 representation of what has happened is next level delusion. Congrats 👏

 

oh hell yea! its weird but AI is kinda random. some days its spot on and some days its unreliable. you have to smack it around a little sometimes.

 

 

 

 

 

 

[xXDeliteXx]

8 hours ago, AvoidTheCrown said:

 

You are right. My screenshots clearly "undermine my own stance" because op tasted the kebab and said "you know what? Let's tune down the spice", thus being the one that made the kebab. You are clearly a desperate person in need to be right so right you are. I hope you get better and if not, you can always ask ChatGPT to help you with it.

Show us your load order.
That's what we want to see.

[flybox]

On 9/23/2026 at 11:16 AM, Z0mBieP00Nani said:

Good point. I never considered that AI could also be used as an interactive tutor.

 

I came back to this thread specifically to comment on this part of the discussion.

 

I have coding experience, but exactly zero of it is in Python and zero of it is in modding. So do I have ideas of mods I'd like to be able to create? Of course. Could I do it? No, not with my knowledge. But do I have the capability of learning how? I think I do, given enough time and effort.

 

So I went back to the AI for another go, and really worked step by step from the very basics to create a "hello world" cheat console command. Each step of the way, I demanded the AI explain to me exactly what I was doing and why, line by line. It took two tries to get it to work (the first time, the AI seemed to be going off outdated code that had changed during a game update at some point), but it worked, and importantly, I understand why it worked.

 

It's now working on teaching me to create a hello world command that opens a notification dialog instead of simply responding in the cheat console. It's a slog. There's an error every step of the way. But each error is new, and importantly I'm learning why the error happened every time.

 

Long story short, this exercise has made me feel more confident about the ability to learn TS4 scripting in the distant future if I stick with it. You just have to make sure you tell the AI to explain everything line by line so you know what you're doing while you're doing it. I even had a moment of learning in this session, when the AI told me to add a certain line to the command definition, and after I added it, I thought, "wait, if I type this here, then I need to add a comma there" and I went back and looked at the entire code that had been generated, and sure enough, there was a new comma. So I naturally started learning a little bit of Python syntax while I was doing this.

 

My main concerns are:

1. Whatever I get to work, there's nothing I have as a baseline knowledge to know if I'm learning how to do things the best/most efficient way. For a hello world script, who cares? But if you aspire to write a longer, more complex mod, how would you know if the AI weren't just helping you cobble together a mess of code that could be redone in a much better way? Even if doing this got to the point where I could say, "Hey, I actually know Python now!" would I have any confidence that I learned at all how to write Python well? Not sure I would.

2. It's extremely time consuming to do this in an environment which requires you to close and then reload TS4 for every tiny change. Unless you are really committed to sticking with it, it might get to be overwhelming very quickly.

 

Anyway, I'm not saying there is a future in TS4 scripting for me, but I will say that this exercise in at least trying has been enlightening.

[fourluvclover]

Remember that nothing you put in ChatGPT is private or secure. Ever. Humans read every single prompt that's submitted, and any information that you feed it will be stored indefinitely. Just food for thought.

[SilentAegis]

4 hours ago, fourluvclover said:

Remember that nothing you put in ChatGPT is private or secure. Ever. Humans read every single prompt that's submitted, and any information that you feed it will be stored indefinitely. Just food for thought.

Eh... this is more a crumb than it is food. Most of this isn't true. OpenAI themselves cite that the number of active users per week exceeds one billion. Back in 2025, when ChatGPT was considerably smaller, users were already sending about 2.5 billion messages per day (this comes to about 29,000 messages every second). So under sheer scale alone, this is pretty implausible. That's a lot of humans to hire just to read people being horny, lol. 

Privacy and security isn't as simple as a on/off switch. Yes, using ChatGPT does increase online exposure, but it does so in the same broad sense that using almost any online service does. Your Gmail, Google search, Dropbox, iCloud, Amazon, banking apps, social media, streaming services... these all increase exposure too with risk that they could be compromised.

The point here is that just because information exists on a company's servers does not mean it's routinely read by employees. Sure, it's true that anything stored online has some privacy risk, but this is a basic risk everybody accepts when they store anything online. You shouldn't paste sensitive info in a prompt, but that's more basic digital-security common sense, and not necessarily evidence that everything you put in a prompt is automatically exposed. 

If online security worked in the all-or-nothing way your claim suggests, then people that work in cybersecurity would either have an absurdly easy job or an impossible and pointless job, lol. 

Edited September 30 by SilentAegis

[Zasz]

11 hours ago, fourluvclover said:

Remember that nothing you put in ChatGPT is private or secure. Ever. Humans read every single prompt that's submitted, and any information that you feed it will be stored indefinitely. Just food for thought.

 

How do you know this exactly? Do you work for the company? Or is this a case of you reading it somewhere on telegram or facebook in groups where people are afraid of their own shadow?

[fourluvclover]

https://www.404media.co/humans-reading-copilot-prompts-images/

 

I'm not trying to tell people how to live their lives, nor am I condemning anyone. You do you. I brought it up because I care about folks being safe.

[SilentAegis]

3 hours ago, fourluvclover said:

https://www.404media.co/humans-reading-copilot-prompts-images/

 

I'm not trying to tell people how to live their lives, nor am I condemning anyone. You do you. I brought it up because I care about folks being safe.

I can respect that. But this article more undermines your original claim rather than supports it. It specifically refers to hiring contractors to review "in some cases" and human reviewers are looking at prompts "sometimes". You took the truth of some prompts being reviewed by humans and inflated it to mean humans read every single prompt and that this happens universally. The former absolutely happens just from moderation alone. The latter is false. It's also not new and the article directly supports that:

"Human contractors being exposed to horrific, unpleasant, or traumatizing material is, of course, not new. Big tech companies, and especially social networks, have used armies of contractors for years to moderate user generated content."

The article also doesn't appear to substantiate anything on conversations being stored indefinitely either. It remains untrue anyway.

Edited October 3 by SilentAegis

[Zasz]

Also the initial claim was about ChatGPT while the linked blog is about Copilot.

[Sheyk]

I am someone who is or used to be almost totally illiterate when it comes to any form of coding. It's only thanks to AI's guidance that I am able to now make small personal edit mods or fixes/tweaks in mods I use. In games like Stellaris or Civilization mostly.

First a disclaimer that I do not know much about AI beside my interactions with them. I have not researched the subject. I use free models so I don't know how much paying for the product would affect the experience.

As an ignorant who spent quite a bit of time working with AI, I am baffled by people who assume you just tell them to do something and they do it. Sure technically it's true if phrased so. But what they did will not be what you asked, they don't do "it", they do something(probably absolute trash). AI's reading comprehension is still abysmal. It provides information with confidence with no regard to the definition of confidence. In essence when you work with AI unless you know exactly how to communicate with it so it doesn't constantly mislead you or forget what has already been established. You will be wasting a lot of time on red herrings to get to a working result. AI's love making assumption and treating them as fact. I just cannot use AI to create databases for me because it won't take the time to look at the data provided, it just make guesses. So in the end I end up wasting more time editing the product it gives me than it would take me to build it from scratch. The compromise I came to is I make them do a baseline format I like and then I edit/fill the terrible product. It might be because I don't know how to use the product properly but in my opinion it highlights the fact that AI is not a magic tool that any unskilled person can use to make a good product without efforts yet.

 

Is the description of my experience the equivalent of going to a restaurant, ordering a meal and then maybe adding a bit of your own seasoning or sauce to the tasty meal you ordered before you get to eat it?
In this analogy the meal is the product. Seasoning or sauce is also a product. When using AI to code you don't just use another code and sprinkle some in it. What you do when you use AI to code or help you code is the process of cooking the meal. You are using a tool to cook. If your goal is to edit the product, the analogy just doesn't work at all. A meal as a product cannot be edited that way. You cannot edit the process that went into making a meal, chemistry doesn't work that way. You cannot change the ingredients used when the original was cooked or the way they were used.

AI is still young. A concept commercially implemented for only a few years is still in infancy. It has a lot of room to grow so it might become what the harsh critics claim it is relatively soon. But right now what critics of AI depict seems to be fantasy not reality. There are in my opinion many negative facets to AIs for general public users right now but most of those negatives exist because they are extremely bad products for general users, not too good. Once they become good then most of those negative facets will be different in nature.

You can probably tell from the way I chose to express myself but using AI for an ignorant is an extremely frustrating process. I am grateful for the opportunity to learn it gives me but it's not an easy tool to use effectively. From my perspective it's a terrible product with the best accessibility.

So I agree with the title of this thread. AI is kinda cool but it's garbage.

Edited October 5 by Sheyk

[SilentAegis]

3 hours ago, Sheyk said:

You can probably tell from the way I chose to express myself but using AI for an ignorant is an extremely frustrating process. I am grateful for the opportunity to learn it gives me but it's not an easy tool to use effectively.

You made several great points I resonate with, but this in particular stands out. 

Over the years, I've found that communicating with an LLM is a skill in and of itself. It's not like texting a person and it should never be. With people, it's so, so easy to rely on tone, implication, social intuition, and all the things that make human communication what it is. 

This all goes out the window with AI. 

Good use isn't just prompting, it also requires supervising. It requires understanding its quirks, both good and bad, and knowing how to steer it to get what you want. It's a capable, but very occasionally unreliable collaborator and getting it to work... well, you don't really think of "watching tutorials on prompting". At least, I never did. You adjust, reiterate, or basically just keep throwing stuff at it until you see what sticks.

The gap between what you want and what it actually gives you depends on both how well you instruct it and how intelligent the model itself is. That gap will probably narrow as AI grows more capable, but I don't think it will ever disappear entirely.

Edited October 6 by SilentAegis

[dooda]

I`m totally with the OP on this. AI is a useful tool for learning how to mod and produce something you want and like. AI does`nt create anything without human input, so crediting a particular AI as the creator is plain nonsense.

 

I`m interested in which AI model the OP has used. I`ve used co pilot for many hours and as it draws on internet knowledge to form it`s mental map it constantly gets S4 coding details wrong because S4 constantly changes over time....BUT... I have learned a lot about how mods work even if I can`t produce the results that people with extensive coding knowledge can.

[flybox]

On 10/5/2026 at 11:19 PM, SilentAegis said:

You made several great points I resonate with, but this in particular stands out. 

Over the years, I've found that communicating with an LLM is a skill in and of itself. It's not like texting a person and it should never be. With people, it's so, so easy to rely on tone, implication, social intuition, and all the things that make human communication what it is. 

This all goes out the window with AI. 

Good use isn't just prompting, it also requires supervising. It requires understanding its quirks, both good and bad, and knowing how to steer it to get what you want. It's a capable, but very occasionally unreliable collaborator and getting it to work... well, you don't really think of "watching tutorials on prompting". At least, I never did. You adjust, reiterate, or basically just keep throwing stuff at it until you see what sticks.

The gap between what you want and what it actually gives you depends on both how well you instruct it and how intelligent the model itself is. That gap will probably narrow as AI grows more capable, but I don't think it will ever disappear entirely.

I think this is spot on. I have resumed my homemade AI-assisted Python learning course, and successfully added a dialog notification Hello World command to my already working cheat console version. It took a lot of trial and error, and especially it required me to constantly feed errors back into the AI, and ask the AI to explain exactly why this version didn't work, and then this one, etc. When I finally got it working, I wanted the AI to explain why it had to go through all this trial and error, and couldn't just teach me the right version in the first place, and the answer seems to be something like this: The AI is really good with Python syntax and mechanics, but not so great with how EA handles Sims-specific details, because it can only build a statistical prediction based on god only knows how much code is out there for it to read, some of which might be outdated or just plain bad.

 

So by taking it very slowly, and expecting full explanations for each failure, I'm not only learning a little bit about how Python works, I'm also learning a little bit about why the AI can't get it right immediately without trial and error.

[Zasz]

Maybe AI has such a difficult time with sims 4 because the code is a mess, unpredictable and totally not what the tool expects based on how its trained.