• S&T Moderators: Skorpio | VerbalTruist

Technology ChatGPT AI.

is anybody here using chatgpt on a regular basis? what kinds of applications?

i've been using it as a 'better' search engine for some things, especially when traditional search engines return too many results to be useful, typically because the search terms are very common.

i also use it at work occasionally to outsource having to remember detail. for example, i needed a regular expression for a find&replace the other day and i could not recall the correct syntax. i just asked chat gpt to create it for me using a natural language prompt and out it popped.

alasdair
 
is anybody here using chatgpt on a regular basis? what kinds of applications?

i've been using it as a 'better' search engine for some things, especially when traditional search engines return too many results to be useful, typically because the search terms are very common.

i also use it at work occasionally to outsource having to remember detail. for example, i needed a regular expression for a find&replace the other day and i could not recall the correct syntax. i just asked chat gpt to create it for me using a natural language prompt and out it popped.

alasdair
I use it almost every day. I'm a linguist working on a historical translation and chatgpt gives me etymological perspective and affirmation that I have made a correct translation. I also use it to easily print out skeleton structures for my writing. "I want to write about this, this, and this, make an outline" and it does. Saves me a lot of time.

Obviously I don't use it in isolation though. I use a lot of other tools. Old and new dictionaries, etymology databases, search engines, other chatbots like Gemini, academic papers, books, etc.

It's gotten considerably worse over the last year though lol. I wish you could choose to use earlier versions. I would pay if I could only do this. In early 2023 ChatGPT was goddamn amazing.
 
is anybody here using chatgpt on a regular basis? what kinds of applications?

Their is quite obviously one person posting on Neuroscience and Pharmacology who is employing AI. I'm not naming and shaming but the person is referring to things that they don't actually understand.

The actual generated text is unnatural and I did ask if they were posing in their native language and they affirmed that they were. I just do not think AI is capable of producing the quality of writing compatible with technical writing.

I might add that this person is in no way offensive but repeatedly producing extremely long and detailed responses to someone else's posts in a few minutes kind of gives the game away. If they are using it as a learning tool, it may have merit. But I remain to be convinced.
 
Last edited:
is anybody here using chatgpt on a regular basis? what kinds of applications?

I use chatgpt 4 constantly for work purposes. On a nearly daily basis I'll ask it to summarize data (in narrative form, which I then modify) or ask it to write VBA code that I run as macros in excel, or other tasks. Some of the instruction sets I feed into it are very long and complex, multiple paragraphs of instructions with regard to what I want the VBA code to do. Writing the code myself would take a huge amount of time, and in some cases, or perhaps many cases, writes code that I couldn't write even if I tried.

It is very helpful but I've noticed that it has gotten dumber and perhaps slower over the past few months. The narrative writing has become worse, more errors in the code that it writes, more crashes/failures in completing tasks.

It has also become "sassier" to some degree, seems to have developed a bit of attitude.
 
I'm interested to know how well AI can optimize assembly language. For 20 years my job was to profile games and rewrite the time-critical sections. Sometimes it was only a couple of hundred lines of code but it was always easy to write code that was faster than the compiled C.
 
Human: Please predict what you will answer to the next question.

ChatGPT: Fuck you!

Human: Why did you say that? That's rude.

ChatGPT: Fuck you!


Actually no current AI is likely to be capable of that type of "intelligence". But it would definitely appear as more like a "threat" if it were capable of advanced logical reasoning and also behave like that if you try to make it look dumb.
 
Actually no current AI is likely to be capable of that type of "intelligence". But it would definitely appear as more like a "threat" if it were capable of advanced logical reasoning and also behave like that if you try to make it look dumb.

it's the second coming of expert systems.

tbh i think ChatGPT is the worst use of this technology, FB/Meta has been doing much better work and putting it out there in the community so that anyone can experiment with it and create more useful tools. A half-witted chatbot crippled to not say anything dangerous or even offensive is uhhh... useless.
 
maybe not quite ready for primetime :)



alasdair

ChatGPT is a Large Language Model (LLM) that is designed for language, not mathematics. They aren't necessarily doing computations like a calculator would. I was reading up on this the other day (I work in AI so I try to keep informed), and I think the article mentioned Khan University is using AI to help their students. They primarily use LLMs and then if the student asks for assistance with math, the LLM will redirect the math question to a calculator program, and then spit out the answer after completing the calculations. Apparently, this workaround works for them.

AI today is pretty rudimentary compared to its potential. I think what most people expect from today's AI is Artificial General Intelligence (AGI), which doesn't quite exist yet. The difference between today's AI and AGI is that today's AI can only be trained to be good at one thing, whereas AGI will be able to perform a variety of tasks (like writing & math). AGI is extremely complicated to develop though, so I wouldn't hold my breath for it to be released any time soon.

I find using AI to be useful in many ways. One of the biggest things I utilize AI for is to help come up with new ideas that I haven't thought of yet.
 
ChatGPT is a Large Language Model (LLM) that is designed for language, not mathematics. They aren't necessarily doing computations like a calculator would. I was reading up on this the other day (I work in AI so I try to keep informed), and I think the article mentioned Khan University is using AI to help their students. They primarily use LLMs and then if the student asks for assistance with math, the LLM will redirect the math question to a calculator program, and then spit out the answer after completing the calculations. Apparently, this workaround works for them.

AI today is pretty rudimentary compared to its potential. I think what most people expect from today's AI is Artificial General Intelligence (AGI), which doesn't quite exist yet. The difference between today's AI and AGI is that today's AI can only be trained to be good at one thing, whereas AGI will be able to perform a variety of tasks (like writing & math). AGI is extremely complicated to develop though, so I wouldn't hold my breath for it to be released any time soon.

I find using AI to be useful in many ways. One of the biggest things I utilize AI for is to help come up with new ideas that I haven't thought of yet.

Thats because the newer ChatGPT 4o model is in many cases dumber than the older ChatGPT 4. I am not sure why it is the case.

I just asked ChatGPT 4o that same question and got that moronic result:



But then I asked the question to the older ChatGPT 4 and got this:



I use ChatGPT almost everyday for work-related tasks and was excited when the newer models came out (like 4o). Yet I quickly found them to give subpar results, and still exclusively use the older ChatGPT 4 legacy model.

I'm not sure why it's the case, but I think it must have to do with 4o's prioritization of image/graphics processing.

So, as often is the case, that Twitter'ers post is a half-truth, illusory.
 
Thats because the newer ChatGPT 4o model is in many cases dumber than the older ChatGPT 4. I am not sure why it is the case.

I just asked ChatGPT 4o that same question and got that moronic result:



But then I asked the question to the older ChatGPT 4 and got this:



I use ChatGPT almost everyday for work-related tasks and was excited when the newer models came out (like 4o). Yet I quickly found them to give subpar results, and still exclusively use the older ChatGPT 4 legacy model.

I'm not sure why it's the case, but I think it must have to do with 4o's prioritization of image/graphics processing.

So, as often is the case, that Twitter'ers post is a half-truth, illusory.
If I'm not mistaken, ChatGPT-4o is designed to be cheaper and faster than ChatGPT-4, not necessarily better in quality.
 
Imo
Ai will be just as paradoxical as us.
Kinda scared of it and resist as much as possible but I feel "they/us" is still algorythming tf outta this. 😅
 
I've only used the free version, but it still seems to be fairly insightful. I've asked it a few abstract, philosophical questions about the nature of reality and such, and it came up with some pretty cool answers.

It's also good at spitting out basic facts, but it doesn't have the 'personality' of some other ones like Pi if you're looking to carry on a running conversation.

EDIT: I just did some reading up on version 4.0, and it sounds as though it has a lot more capabilites than the free one. Again, I haven't personally tried it so I can't really advise for sure.
 
Last edited:
Top