Reflections on an Early Experiment in Chess Programming: Drawing Parallels to Modern Machine Learning

Looking back over the years, it’s fascinating to see how small moments, seemingly insignificant at the time, can resonate with the major technological shifts of today. In 1989, as a cocky 20-something, I made a bet with friends that I could write a chess program in a month. This was the year Garry Kasparov famously defeated IBM’s Deep Thought, sparking discussions about computers, chess, and artificial intelligence. My goal wasn’t to revolutionize technology—it was simply to win a bet and sharpen my programming skills.

At the time, I was an Analyst Programmer on the IBM System 38, working in RPG (Report Program Generator, for those who might confuse it with Role-Playing Games 😊). Without the internet, I relied on the library, where I found a book on chess programming. Though I’ve forgotten its title, the ideas it contained still feel remarkably familiar in the context of modern machine learning and reinforcement learning.

My Chess Program: A First Experiment in Logic

The chess program I wrote wasn’t groundbreaking, but its structure now feels oddly prescient. It included principles that, in hindsight, resemble the foundations of machine learning:

• Board Evaluation: I assigned values to pieces and positions, calculating the board's score for every possible move. Positive or negative scores indicated the relative strength of each position.
• Search Depth: The program evaluated moves three layers deep, identifying the top-scoring options at each level to determine the most advantageous sequence.
• Learning from History: It recorded moves and their outcomes in a file (Physical Filefor System 38 veterans, a "table" for everyone else). This allowed it to reference past results, avoiding repeated mistakes and reinforcing successful strategies.

Board Evaluation : I assigned values to pieces and positions, calculating the board's score for every possible move. Positive or negative scores indicated the relative strength of each position.

Search Depth : The program evaluated moves three layers deep, identifying the top-scoring options at each level to determine the most advantageous sequence.

Learning from History : It recorded moves and their outcomes in a file ( Physical File for System 38 veterans, a "table" for everyone else). This allowed it to reference past results, avoiding repeated mistakes and reinforcing successful strategies.

The approach was rudimentary but conceptually aligned with reinforcement learning: evaluate, predict, act, and learn.

What Worked, What Didn’t

The program could move pieces, capture opponents, and navigate the board with basic logic. However, it lacked a comprehensive understanding of chess and couldn’t aim for an endgame victory. It was never meant to rival Deep Thought, and life soon pulled me in other directions. But revisiting it now, I see how those simple steps mirrored foundational concepts in machine learning.

The Machine Learning Filter

Fast-forward three decades, and we live in a world reshaped by the digitization of everything. Machine learning is now at the core of how businesses, governments, and individuals operate. Data modelling, prediction, and learning loops are the forces driving innovation. My old chess program, with its attempts to assign value, predict outcomes, and learn from history, was a tiny precursor to the immense power we see in AI today.

Here’s the truth: If you’re not making machine learning part of your everyday toolkit, you’re likely wasting time and energy. The efficiencies and insights it offers are game-changing—but here’s a word of caution: always sense-check the output. Machine learning can be dazzlingly convincing even when it’s wrong, so human oversight remains critical.

The Future: A Question of Dominance

As we embrace this AI-driven future, I can’t help but wonder: where will we be in five years? Will Google continue to dominate search, or will a machine learning rival dethrone Chrome as the browser of choice? Change is inevitable, and as the pace of innovation accelerates, new leaders will emerge. What remains constant is the need for curiosity, experimentation, and the willingness to adapt.

My old chess program was a small experiment fuelled by youthful confidence, but it planted seeds that echo today. It’s a reminder that even the simplest projects can offer insights into how we think, learn, and solve problems. Machine learning isn’t just a buzzword—it’s the future. Embrace it, explore it, and, most importantly, always stay curious.