ChatGPT and the Art of Creative Problem Solving

Unlocking Our Coding Potential with AI-Enhanced Freedom

Many of us have witnessed or heard of countless instances where ChatGPT made ridiculous mistakes, or heard claims that AI will replace developers. I'd like to share a couple of recent experiences.

When it comes to technology, I have always resisted learning skills that are specific to a particular library, language, or platform, as these tend to become outdated very fast. For instance, I don't see the need to still memorize calls to libraries like LIBGPU or LIBGTE for PS1 (PSX), how to provide MS-DOS access to more than 640KB of RAM (EMM386), or the machine language program for erasing all memory on the Sharp PC 1251 (0x26 0x2D 0x02) (just in case my teacher discovered my calculator contained cheats). Even the best practices for programming languages seem like unduly complex and ever-changing trivia to me.

In contrast, skills such as math and algorithms are intrinsic and remain useful over time, regardless of technological changes. Knowledge of linear algebra, signal processing, and fundamental computer architecture concepts have remained valuable across multiple technological disruptions.

What is remarkable about ChatGPT is that it is exceptionally proficient in extrinsic knowledge. By parsing thousands of lines of code, it can accurately use a multitude of libraries and interfaces. This means that all the effort put in by hundreds of programmers to learn how to call an interface is now available to others. There is no need to sieve through GitHub or StackOverflow for answers when you are stuck.

In the past few weeks, I have had two enjoyable experiences. I wrote a small Python program using PyQt5 (without knowing anything about it and without reading any documentation) and assembled a demo iOS SwiftUI app. In both instances, I was able to get up and running quickly, without reading documentation, to the point where I could write the code I wanted in no time. This is revolutionary to me.

There are certainly limitations to ChatGPT even with this type of use. For instance, I encountered an issue where there are multiple web3 Swift libraries, all with the same name "web3.swift," and ChatGPT produced one integrated "fusion" dish. In general, however, most minor code glitches generated by ChatGPT can be resolved either by asking it to correct itself, which it does with apologies, or by using common sense. Overall, the amount of time and pain saved is incredible.

What's even more intriguing is the recent progress ChatGPT has made in answering complex questions. The transition from version 3.5 to version 4 has been astounding. For instance, I asked the question:

How can someone prevent front running while minting a unique NFT from a QR code?

Version 3's response was nonsensical and irrelevant, merely rephrasing the question. In contrast, version 4 provided specific and practical suggestions and offered multiple valid solutions to any objections raised.

In areas of with innovation or invention, the creative process is founded on recognising a problem as an opportunity rather than an unchangeable fact of life. Daring to ask the question is generally the trigger, the invention that follows is about standing on the shoulders of giants and leveraging prior art, not about flashes of genius. ChatGPT makes the shoulders more accessible to all and the work of those who precede us more useful.

Could the human touch be in asking the right question, crafting a prompt that unveils a valuable insight?

#innovation #creativity #assistant #chatGPT #AI #developer