ChatGPT and Uniface: Friends Or enemies?
Software Imaging
With a worldwide reputation for bespoke application development we are leaders within the technology field.
With the rapid evolution of technology, AI chatbots like OpenAI's GPT have been revolutionising how we interact with computer programming. They've proven to be highly useful for scripting in Python, HTML, JavaScript, and even for creating complex SQL queries. These AI models can efficiently generate code from a brief description of what you need, acting as a virtual programming assistant that saves both time and effort.
However, when it comes to Uniface, a high-level programming language used in enterprises for building mission-critical applications, AI's proficiency remains a bit ambiguous.
In recent experiments with GPT-3.5, attempts to create Uniface Procscripts have yielded less than satisfactory results. The code produced often does not reflect the familiar syntax and structures of Uniface's proprietary Procscript. The generated scripts required considerable manual adjustments, effectively serving as pseudo-code rather than fully functional Procscript.
A subsequent test with GPT-4 showed improvements in code generation, suggesting that subsequent iterations of AI models may better learn and emulate the syntax of less common programming languages. However, the validity of the generated Uniface code remains unconfirmed due to the limited opportunities for thorough testing.
Given these findings, the key question emerges - Should Uniface professionals consider collaborating with AI companies to train their models specifically on Uniface?
Such a partnership could unlock immense potential. A machine learning model trained on extensive Uniface datasets could become a powerful tool for Uniface developers, enabling faster code generation, increased productivity, and enhanced code quality.
In summary, while the current generation of AI models like ChatGPT may not fully replace Uniface developers or eliminate the need for human touch in Procscript writing, they hold promise for the future. The advancements in AI and machine learning might not be far from understanding and generating code in more specialized programming languages like Uniface.
If you have experiences to share regarding using AI for Uniface, please share them in the comments below.
.Net Backend developer
4 个月And for that I decided to learn uniface
?DevOps Engineer | ?Cloud Engineer | ?Software Engineer | ?Azure | ?.NET | ?System Integration | ??Cyber-Security
1 年You need to train GPT to do it in the first place. An hour of teaching session through conversation will get you surprising results. You'd see huge improvement in accuracy of code generated by the AI, within context/scope of your thread/conversation. Getting the knowledge pulled into the core model is a different story (making it available to other users). But keep in mind that it can still leak your code, so best to ask it to write small bits of code and then put it together yourself. Lastly, if it gives you faulty code, correct it yourself and send it back with explanation so it can learn from you. You can also feed it links do documentatiom pages (in some cases it will get some data out of them), alternatively you can feed it chunks of documentation in plain text. Anyways, I successfully trained it do a lot of cool things in code and it wrote great code that worked straight away based of about 20 different requirements. (10 classes in one response)