Install Steam
login
|
language
简体中文 (Simplified Chinese)
繁體中文 (Traditional Chinese)
日本語 (Japanese)
한국어 (Korean)
ไทย (Thai)
Български (Bulgarian)
Čeština (Czech)
Dansk (Danish)
Deutsch (German)
Español - España (Spanish - Spain)
Español - Latinoamérica (Spanish - Latin America)
Ελληνικά (Greek)
Français (French)
Italiano (Italian)
Bahasa Indonesia (Indonesian)
Magyar (Hungarian)
Nederlands (Dutch)
Norsk (Norwegian)
Polski (Polish)
Português (Portuguese - Portugal)
Português - Brasil (Portuguese - Brazil)
Română (Romanian)
Русский (Russian)
Suomi (Finnish)
Svenska (Swedish)
Türkçe (Turkish)
Tiếng Việt (Vietnamese)
Українська (Ukrainian)
Report a translation problem
In order to properly instruct chatGPT you need to know lua and its specific application in SW, with all of its intricacies.
If you really do I recommend learning what LLMs are good for from someone like Yann LeCun, he's more expert than anyone here.
In short Large Language Models are good for manipulating text, that's it, so for example if you wanna do a reverse dictionary (finding a word based on its definition) , chatGpt might be good for that.
SO yeah if you already
some LLMs might be good assistants to go a bit faster (don't know about chatGPT specifically though, might not be), but that's about it.
What LLMs are absolutely not are epistemological machines, they have no knowledge, they have no reasoning, they can't code something for you based on a description of what you want because they don't have any conceptualization of the world.
ChatGPT is especially horrible at that since it was made for laypeople and made to trick people in making them believe it actually does understand, when it absolutely does not. It gives you the impression of knowledge because you can ask him for simple things and it will regurgitate a predicted text based on text found on the web, but it is meaningless text, literally meaningless, it can't reason, it cannot differentiate between something true and something false.
But not Stormworks specific lua.
Someone tested it with MIPS and it apparently did an excellent job at that too.
And thing is you could actually teach ChatGPT to understand Stormworks Lua.
So MAYBE if we provide more data to it would it be able to do more stormwork lua soon?
At least basic steering and braking sound promising
Also make it analyze actual stormworks lua scripts see if it can disect it as it does it more it may learn from it.
So you don't actually want to learn from experts right ? You just ask random people here and trust their answers based on .... ?
Anyone telling you that chatGPT can learn to code in LUA for Stormworks doesn't know what he's talking about.
First reason being that they have updated many times chatGpt to prevent it from learning things, because since it cannot have knowledge as we understand it as humans, but only predict seemingly correct sentences, people can teach it abhorrent things pretty quickly.
So they made it so it cannot learn from users, they also heavily moderated its output to ensure it will not say politically deviant things, they're not about to let it "learn" from non-approved sources.
https://www.theatlantic.com/technology/archive/2022/12/chatgpt-openai-artificial-intelligence-writing-ethics/672386/
And here is Yann Lecun explaining what LLMs are :
https://youtu.be/ULbpPHjiSBg
Where our privacy and security is important and matters to them so much.
When they are the ones who infringe on you constantly.
I don't know but if I want to make a point last thing I would do is use a META post or a statement from META or any of it's employees.
That said, i use this tool alot, because it's way faster than going in and out of your creation non-stop:
https://lua.flaffipony.rocks/
I'm no AI or Lua expert, but it would be pretty easy to test this for yourself if you're that skeptical. I was able to replicate the basic function of my own braking controller, but things began to fall apart once we got into traction control. For steering, it was capable of ackermann with inputs for wheelbase and track, but again, once we got into speed sensitive steering, it really seemed to fall apart.
This was in GPT-3.5 I will try again and see if GPT-4 is any more capable.