My personal story
I'm a 75-year-old retiree whose entire career was spent immersed in software and computer hardware - yet I never wrote a single line of code myself.
Upon retiring, I became captivated by the rapid growth of Artificial Intelligence. Is AI truly the next industrial revolution, or simply advanced hype?
Will human expertise become obsolete?
These fundamental questions led me on a mission of discovery: to prove what AI is capable of today and what still remains out of reach.
I've always understood classical computer logic: the path-dependent 'if/then' flow. This logic offered no reasonable explanation for how a machine could recognize a handwritten digit with near-perfect accuracy.
My AI Deep Dive
I committed myself to building a handwritten digit recognition program from scratch. With only the free versions of ChatGPT and Gemini as mentors, I learned Python and followed their guidance — step by step - to bring the project to life.
It took six months of intensive work, including revisiting high-school math concepts like multidimensional functions, derivatives, and vectors.
The Outcome: I successfully built a functional AI model with a 93% accuracy rate - a remarkable achievement for a coding novice.
Through this process, I learned the truth: AI "learns" by adjusting thousands of mathematical parameters—784 per image across 20,000 samples in my case—to find the best-fit multidimensional function,
This deep-learning process is why modern AI demands immense computing power and the parallel processing capabilities of GPUs.
What I Learned About Working with AI
- AI will Redefine Programming Roles: AI writes accurate, functional code in seconds. It will rapidly replace entry-level and routine programming tasks.
- Human Command is Crucial: AI tends to over-deliver, offering a deluge of details and options. While some advice offers brilliant insights, you must maintain absolute focus on your core goal to prevent scope creep.
- Manage AI's Limitations: If the AI goes off-course or gets stuck in a conversation loop, don't attempt to debug the thread restart the process immediately.
- The Architect is Human: AI doesn't grasp the entire project architecture. The developer must break down large projects into clear, small modules, feed them to the AI one by one, and meticulously verify that the output meets specifications.
- AI is an Unbiased Debugger: Skilled programmers can sometimes be blindsided by their own assumptions. AI often has an unbiased view of the whole programming environment and can fix non-obvious bugs where human experts least expect them.
- AI is an Advanced Search Engine: Fundamentally, AI is a sophisticated search and presentation tool. It does not "understand" the business context. You are the captain; success hinges on the clarity of your instructions and your final judgment.
You are the captain; success hinges on the clarity of your instructions
