ChatGPT-4 can do Purebasic, here is your Prompt.

[TheoG]

I case you haven't tried.
ChatGPT-4 can meanwhile do Purebasic very good. However, I must say that you have to pay the $20 and get the PLUS Version.
The free version might be disappointing at some points.

So we are talking from GPT-4.
If you do only use it from time to time, you can try this: Open Playground.

I will give you a prompt that you can use as a starting point:

Code: Select all

Act as a Purebasic and JSON Expert. Continue this Subprogram:
 Procedure.s UnescapeJSONString(Est.s)
  Protected Une.s, i.l, Ch.s, nc.s, Hx.s, Uc.s

Just enter it and be amazed.
Of course you can improve this Prompt, and i want to give you a Sample on how far you can go.

This is just for fun. And to make you experiment with it. I call it the "Split Personality Prompting".

Code: Select all

 I want you to act as 10 Purebasic developers,  "A", "B" to "J".
 All these Purebasic Developers are experts in different fields of Purebasic Code and x86 Inline Assembler, show me their fields of expertise.
 I will give you a task to solve that all ten person will solve, but you will not output their result,
 just output "Solved A","Solved B" until "Solved J" together with a short comment what is special with their solution.
 You will then internally let an Expert Admin evaluate all 10 results,
 compare them and then combine all 10 results to one perfect result that you will output.
 Here is the task to solve: "Make a Purebasic Program to  to get the number of Users out of the Active directory"

There is another new Competitor on the AI-Market however it can only be used from US and UK.
So i could not test it. But its very promising:

Claude 2

If you are from GB or US, maybe you can try if it speaks Purebasic already.

[Bitblazer]

If you are from GB or US, maybe you can try if it speaks Purebasic already.

You could use a VPN.

ps: other models like MosaicML's MPT-3b can do purebasic too, but the quality is ... medoicre.

[Rinzwind]

Claude: Admit when I don't know something and try to be honest about my capabilities

Already better than ChatGPT

[TheoG]

Hello everyone,

I'm excited to share some intriguing insights based on my recent experiences using AI, specifically GPT-4, in conjunction with PureBasic and PowerBasic programming.

Over the past few weeks, I've noticed a clear improvement in GPT-4's proficiency with both PureBasic and PowerBasic. This 'learning' curve could be my own growing familiarity with the tool, but I'm convinced that the model itself is also improving week by week. This suggests that even though these AI models officially had their training cut-off in September 2021, their utility and accuracy are still on the rise.

At present, GPT-4 is a valuable companion when working on single procedures, offering insights and generating code that has greatly assisted my workflow. While I haven't had a chance to test Anthropic's Claude yet, I anticipate that with enough user interaction and correction, it could soon excel in generating more substantial PureBasic programs.

One aspect of this evolving AI capability that particularly fascinates me is inline assembly (Inline ASM). Despite its initial uncertainty, GPT-4 can handle this too. You can provide a procedure to the AI and request it to "complete it as Inline ASM". Admittedly, this area is still developing, and the AI doesn't always hit the mark. However, I have had some success, even when working on complex sorting algorithms like the Boyer-Moore algorithm.

As we continue to explore the synergies between AI and coding, I encourage everyone to share their experiences and findings. Your input could help fine-tune these AI models and broaden their capacity for coding assistance, whether it's PureBasic, PowerBasic, or other languages. Our collective expertise and experiences will drive these AI models to new heights, ushering in a new era of AI-assisted coding.

[DK_PETER]

I truly hate AI and I can't stand the owners and the developers creating it..

I don't expect many to join me in this, but the future they want
AI to participate in and what it will be used for is not good.

Already it is biased as hell. Don't ask me what I mean - try it and
you might find out yourselves. When you zero in on a specific subject
you'll find out that it is set to follow a predefined narrative.

I deeply believe that it will become a severe leash around our necks - not a creative tool that will
set people "free".

Just my two cents. This will be my only comment on this subject.

Have a nice day.

[HeX0R]

I completely sign your whole statement!
Another few of those damn Chat-GPT, superduper-GPT or most-unnecessary-shit-in-history-GPT threads here and I probably quit programming and become a beekeeper.

[yuki]

@OP: #1 + #2 + #3 are all fairly large blurbs of text, posted within a rather short span of time. The extreme level of similarly oriented marketing-speak feels potentially insincere and definitely suspicious to me.

As someone who works with similarly overenthusiastic people: I'm sick of fixing broken code where the AI is blindly trusted. It's not ready for primetime and even when it becomes so, it's based on misappropriated IP. (source: trained on StackOverflow data but appropriate attribution is not given)

I truly hate AI and I can't stand the owners and the developers creating it..

I don't expect many to join me in this, but the future they want
AI to participate in and what it will be used for is not good.

I wholeheartedly agree.

Considering the current trajectory, the human component will inevitably be minimised, with large firms leveraging AI to both discover ideal concepts and generate content across various mediums. Large media firms and social media companies already have teams working towards this goal.

With individualised content-generation satisfying the individual's desire for effectively constant media consumption today, legitimate creative works and innovation will be drowned out, as people simply ask the AI for tailored content.

Then, the only winners are companies like OpenAI raking in subscription $$$.

Hustlers selling "prompt engineering" courses will disagree while claiming you can get super mega bucks superfast by just feeding the AI simple concepts and churning out apps.

Sorry for doomposting.

Another few of those damn Chat-GPT, superduper-GPT or most-unnecessary-shit-in-history-GPT threads here and I probably quit programming and become a beekeeper.

Hey, any more of those beekeeper positions open?

[Little John]

@OP: #1 + #2 + #3 are all fairly large blurbs of text, posted within a rather short span of time. The extreme level of similarly oriented marketing-speak feels potentially insincere and definitely suspicious to me.

Same for me.

[Fangbeast]

While I won't piss on anyone's opinion and they have a right do doom themselves if they wish, I truly think that I don't want to be a tube fed, drooling idiot that is essentially a dead meat sack that can't do anything or think their way out of a paper bag because ai is doing everything for them, at least not until I am a stuffed up geriatric that has no choice (heheheh). SHADDAP IDLE!!

To hell with AI, I will think for myself and am currently registered for wasp keeping zoos in South Venezuela. Bloody hell they hurt when they sting though!

[Fred]

@OP: #1 + #2 + #3 are all fairly large blurbs of text, posted within a rather short span of time. The extreme level of similarly oriented marketing-speak feels potentially insincere and definitely suspicious to me.

Same for me.

Agreed

[jacdelad]

The last time I spoke against AI in this forum I was told I'm being misinformed and do not understand the topic. Interesting...