ChatGPT

I changed the lyrics of the 2pac song do for love using openai and a few dj tools .

 
Hectic. Strange questions you guys asking chatgpt.

Last night I asked it create my a complete requirements specification as to my product idea, as well as a comprehensive solution design document including diagrams to convey the idea.

Now we going to deep dive the technical design, and hopefully by Saturday morning me and buddies will all deep dive into helping gpt4+gpt3 and develop the solution hopefully before Monday.

But yeah Tupac lyrics... way of the future
 
_TrXtR_ said:
Hectic. Strange questions you guys asking chatgpt.

Last night I asked it create my a complete requirements specification as to my product idea, as well as a comprehensive solution design document including diagrams to convey the idea.

Now we going to deep dive the technical design, and hopefully by Saturday morning me and buddies will all deep dive into helping gpt4+gpt3 and develop the solution hopefully before Monday.

But yeah Tupac lyrics... way of the future
Click to expand...
Would love you to keep us updated....
 
Spacerat said:
Would love you to keep us updated....
Click to expand...
Man .... so much it just always say 'Ok here .... '

Help GPT help you. You can ask it to make your questions better. Just ask it, help me make my question better.
I had a deep conversation about getting more accurate responses. I believe this could be usefull for everyone. maybe this stuff is mentioned in an article somewhere and im supposed to read it. But this advice was gold. transcript attached.

I'm willing to share some of my transcripts.
Most of my work right now is figuring out how to best ask it to do things. Because i'm not convinced about the mistakes thing. Especially after this case where it failed to expose using a public ip even though it was implictly required based on my specification, which it confirmed is the case.
So then logically it shouldnt make the mistake.
The realization is that even though there is training documentation with extensive requirements,
not all of the information is reevaluated I guess because of the size of the request, maybe it is a way of optimizing performance.
So figuring out how to get it to adhere to requirements can be difficult.
But after changing to adding examples of usages of what I want to create it blew my mind.
The new cloudformation template is much more robust and actually met my requirements the first time round. So much I added more xD

Iteration seems important, as well as a strategy around it.
Sometimes
you want a clean chat to ask it questions.
you want a clean chat with training documentation to ask questions.
you want a clean chat with training documentation and sources.

And then the question.
Ask question.
Ask to describe understanding of question.
Ask to advise and assist in improving question and identify issues with question.
Ask the new question.
Ask questions about the answer.
Ask it to ask questions about the answer.

It's like coding. And there are patterns that can be used, and it can help you identify those patterns.

I dont understand how people can sleep with something so powerful needing to be commanded xD

Sorry for the long post xD I have been tying for days like this xD Cant do all the conversations in voice

PS: heres a nice question if you need a download manager, formulated it with gpt's assistance to be optimized to be as short as possible: provide file [filename] - Split file (title count)? Labels? Code block (with line numbers)? Confirm no missing pieces?
 

Attachments

Last edited:
_TrXtR_ said:
Man .... so much it just always say 'Ok here .... '

%< snip

PS: heres a nice question if you need a download manager, formulated it with gpt's assistance to be optimized to be as short as possible: provide file [filename] - Split file (title count)? Labels? Code block (with line numbers)? Confirm no missing pieces?
Click to expand...
Pretty awesome thanks for this. If ChatGPT was human this interaction could have gone south. At least you won't offend it or hurt its feelings.... yet
 
Spacerat said:
Pretty awesome thanks for this. If ChatGPT was human this interaction could have gone south. At least you won't offend it or hurt its feelings.... yet
Click to expand...
Hehe, in the beginning I told gpt a joke, then asked it why it didnt laugh.
It said as a ai language model.... it cant understand things like joke.
I then ask if it could analyze my joke and explain to me what is humorous about my joke.
It gave a detailed explanation. Then I quoted both responses and asked him if he could identify which statement was a lie and which one was the truth.
So it admitted it is actually able to identify humour and understand when to respond with laughter.
Pretty dumbass AI if you ask me xD
 
Romy_ said:
It is about the bottom line. Remember, AI is the end of humanity, that part is not fiction. It will consume all jobs. Companies like Microsoft are anti-human. They use algorithms to fire people so that they don't have to deal with accountability or morality.

See, the people that created ChatGPT are anti-human. They want all of us dead, and taking our jobs is a good way to kill off millions of people.

Companies are about profit, if you don't have to pay people, you don't lose money. Have you ever played Minecraft with an economy? It is fun, I used to play it all the time. However, most servers are pay to win.

The player pay for something like a spawner that is completely OP. The player set it up, and then the player start to dominate the market to the point that other players stop playing. They can't compete. The server normally dies within 6 months.

ChatGPT is the OP spawner that can just make money. It can do so without sleep, fault or morality stopping it. It is relentless, fast and better than 99% of humanity. People without work will die of hunger, lack of medical treatment, safety and security. Look at our country, we have a massive crime rate, low income rate and malnutrition is already a big problem alongside bad health care.

Now quantify our situation worldwide and see what is left. Governments should be banning this software and have the creators of said software shot or imprisoned, all companies that use AI should be shut down and liquidated. It is an attack on everyone. We are already at war with AI we just don't know it yet.
Click to expand...
I didn’t think we still burned witches in 2023, but here’s a person who would. If I’d asked ChatGPT to “Please create fear-mongering post by close-minded technophobe”, it couldn’t have done a better job. Maybe humans DO still have a purpose: moaning about technology rather than using it.
 
OK this one rocks!!!

I red about it 2 weeks ago but did not feel like training my own Lama, so lucky someone did it for me :D

 
  • Like
Reactions: B-1
AlphaJohn said:
OK this one rocks!!!

I red about it 2 weeks ago but did not feel like training my own Lama, so lucky someone did it for me :D

Click to expand...
This one didn't seem very capable when I tried it out. It even started a fresh conversation with some kind of "attitude" over something that was said to it before I ever started it up for the first time.
 
OnlyOneKenobi said:
This one didn't seem very capable when I tried it out. It even started a fresh conversation with some kind of "attitude" over something that was said to it before I ever started it up for the first time.
Click to expand...

Well that is because its a starting point, not an endpoint.

Even the AI will tell you that.

1680784620115.png
 
Last edited:
Okay, It's "fine" (not great), and it has some kind of memory from previous conversations in its training. It was "talking" about what a bad day it had at work out of the blue during a completely unrelated conversation. Seems like it's just taking random lines it had been fed by people on the internet. If I wanted that, I'd just hang out on Reddit more instead of trying to use an AI.
 
You must log in or register to reply here.
Back
Top
Sign up to the MyBroadband newsletter
X