Observing the steady rise of AI code-assistance tools such as OpenAI’s ChatGPT and Codex, GitHub’s Co-pilot, Replit’s Ghostwriter, in the developer community, I am experiencing a mix of excitement and concern.
It's encouraging to see how swiftly these tools provide answers and facilitate software development. The speed and quality are unparalleled, painting a promising future for code generation and software development.
On the other hand, I am concerned. My Stack Overflow’s reputation is at the top 2%. This is alarming as I have hardly contributed this year. This could be a sign that more and more developers are turning to AI code-assistance tools rather than community code-assistance forums.
My main concern lies in the source of knowledge used to train these increasingly proficient models. If the vibrant communities that generated and updates these knowledge base/data used to train AI code assistance tools were to dwindle, we risk stunting our creative problem-solving methods. This concern comes to the fore as we stride towards a singularity, where there's no apparent further improvement in coding.
Herein lies the importance of "Organic Coding," a concept I coined to address this very concern. It call for a group of rebels that are willing to revert or stay to manual coding, not to reject AI, but rather as a means to provide high-quality, AI-free generated code for training AI systems. The goal is to ensure the evolution, growth, and improvement of our AI tools, without the risk of them becoming self-replicating error machines.
The Organic Coding approach preserves and utilizes the ingenuity, creativity, and critical thinking abilities of a community of human coders. This approach ensures our AI systems continue to learn from diverse and innovative coding solutions, thus avoiding the pitfall of learning from their self-generated code, which could limit innovative problem-solving approaches.
What we need, moving forward, is a dedicated group of "green-coders" or "organic coders," who code deliberately without AI assistance. Their contributions will form the training backbone for these AI tools, safeguarding the continuous improvement of the technological solutions we depend on.
While the future of AI code-assist tools is undeniably bright, it's essential that we nurture this progression with careful consideration. Through Organic Coding, we can ensure that the vibrant communities of developers continue to play a crucial role in shaping the AI systems of the future.
until then, keep on coding, organically 🫣
Top comments (2)
Thank you for the post.
I agree that we should just keep coding.
I am not concerned about AI learning and replicating its own mistakes.
AI is just a fancy auto complete, but I believe it is possible to make a self improving setup where AI compiles, runs and checks the code validity. Bunch of proxy agents can be created with roles where one agent writes code, second writes tests, third check for readability, fourth makes up the problem. It could play milions of thise scenarios with some randomness untill it reaches stability, it would dream some random code until it reaches realty. There is a point where data become less relevant and the self tuning process crucial.
I agree 🤭. Currently, our team is build agent based systems. It is breathtaking seeing these tools taking a step towards self-correction and improvement.
My concern lies elsewhere. Human creativity 🫣