Conversational Programming: How co-creating code with AI's is already a thing

Recently I started exploring a new feature of ChatGPT: the GPTBuilder. It's available to Plus subscribers and is a mechanism to create your own, personalized, and specialized GPT "apps" that will be publishable in the near future. At first, I wasn't sure if my Python skills were up to snuff so I shied away from it. After thinking about a bot I'd like to try creating, however, I decided to give it a go.

The realm of software development is undergoing a transformation, one that's as significant as the leap from assembly language to high-level programming languages. This transformation is conversational programming, and this was my first go at working with an AI to develop another AI without drafting a single line of my own code.

The Traditional Programming Paradigm

In traditional programming, the rigidity of syntax and the unforgiving nature of compilers have been a staple. How many of us spent HOURS searching for a bug in our C++ code just to find that we forgot a ";" somewhere? Coding and debugging is a methodical process, sometimes tedious, that demands a meticulous understanding of structures and algorithms—a process described by veteran programmer John McCarthy as:

the meticulous manual of a digital craftsman.

The Conversational Programming Shift with GPTBuilder

GPTBuilder (and there will be other 'builders' in the future!) offers a stark contrast to this experience. "Programming" with GPTBuilder is more akin to teaching a person how to perform a series of tasks. It starts by sitting that person down and giving them context for what they should learn. I have a background in Tarot, Astrology, Numerology, and the Tree of Life (all of which I'm happy to discuss in greater detail). I had to start by telling the bot what kinds of info to gather, who are some trusted sources that I personally know for it to learn and synthesize information from, and how should it think about combining and intersecting those topics in ways that mimic my own unique style. I added some "trademark" Matt Brooks techniques by describing to it just how I look at, combine, and consider certain symbols and patterns.

Each time I taught the builder something new, it would diligently enter a 'coding cycle', confirming to me at the end that the GPT was updated with a particular skill. On the right side of the window was a 'playground' for me to test out what I had just done. If something wasn't quite right, I could give updated instructions asking the builder to emphasize certain things or avoid others. I was able to come back into the builder days later after some real, road tests, and provide similar updated instructions to tweak and re-publish the GPT app. This feedback and update loop is completely different than traditional coding and takes a few seconds instead of several hours.

The Joy of Interactive Coding

This new way of interacting with AI to program is not just innovative; it's downright enjoyable. The AI understands my objectives (even if it takes a couple of tries to explain them to it), translates them into code, and even suggests improvements and new ideas. As AI pioneer and expert Kai-Fu Lee puts it,

AI is taking us from a world where we have to understand computers to one where they will understand us.

Conversational programming with a builder AI doesn't just change the "how" of programming; it changes the "who". It opens the field to those who may not have formal coding expertise but possess strong problem-solving skills and clear communication. The focus shifts from syntax mastery to conceptual clarity and design thinking. There is a whole skill set to hone that's simply focused on how to talk to AI bots in a way that helps them to understand clearly what one's vision is and then let them change that into code. Just like in any relationship, communication will truly be key!

Creativity Unleashed

The most striking aspect of GPTBuilder is the freedom it affords. I'm no longer bogged down by the tedium of debugging and searching for semi-colons. Instead, I'm liberated to focus on the creative aspects of designing and creating, on innovation, and on tuning the outputs to see my vision come alive. As software engineer and author Ellen Ullman has observed,

We build our computer systems the way we build our cities: over time, without a plan, on top of ruins.

With GPTBuilder, we have the chance to change that—to plan and build with intention and insight.

?? Into the future we go!

Programming with GPTBuilder isn't just a novelty; it's a glimpse into the future of technology—a future where our conversations with machines are as natural and productive as those we have with our colleagues. If you're excited about the future of tech, or just curious about AI's potential, it's time to explore conversational programming, and I recommend starting with GPTBuilder if you're a ChatGPT Plus subscriber!

?? In the meantime, check out MO, my first GPT app! He's ready to help provide guidance on any questions you have. Below is a screenshot of a conversation I had with it. What's especially cool about him is that MO can interface with Dall-E to generate a metaphorical picture of your readings!

Footnotes

1. McCarthy, John. "Recollections on My Career in Programming." Computer History Review, 2012.
2. Lee, Kai-Fu. "AI Superpowers: China, Silicon Valley, and the New World Order." Houghton Mifflin Harcourt, 2018.
3. Ullman, Ellen. "Life in Code: A Personal History of Technology." MCD, 2017.

Matt Brooks is a seasoned thought leader and practitioner in data and analytics; culture; product development; and transformation. View additional articles by clicking below; like them/follow him to learn more!