I can't recall which Christmas it was—likely 2004. The standout memory from that day is the disappointment I felt when my mom gifted me "Artificial Intelligence: A Modern Approach" by Peter Norvig and Stuart J. Russell.
Who Needs an Xbox When You Have an AI Textbook?
Priced at $115.00 AUD, it wasn't a cheap gift. But focusing on a dense textbook is challenging when everyone else is flaunting their presents. I didn't crack it open until I began a job in the city, which came with a 90-minute commute. With earphones in and neutral music playing, I finally delved into the book, almost cover to cover. This book was a game-changer for me in several ways. First, it made me realize that groundbreaking technologies are built on foundational principles. AI essentially creates a multi-dimensional space of potential solutions, iterating to find the best one for the task. I blogged about this, gaining traction on HackerNews. A UK grad student even reached out for collaboration, which I declined due to imposter syndrome. Second, I understood that the field wasn't beyond my intellectual grasp. Despite the jargon and complex terminology, the core concepts were accessible. I even started generating original ideas in AI.
However, I soon got sidetracked, focusing on bridging the gap between academia and practical skills. In 2011, Coursera launched, offering AI and ML courses by industry leaders like Peter Norvig and Andrew Ng. I enrolled and completed both. Eager to apply my newfound knowledge, I turned to Python's robust data analytics libraries, such as PyTorch and NumPy. However, I found that industry practices were far more abstract than I had anticipated, leaving me disoriented. Fast forward ten years, and I've mostly abandoned my AI aspirations, settling into a stable career as a "conventional programmer."
The New Wave of AI
This AI is different—convolutional neural networks that back-propagate and measure deltas across layers. Access is limited, mostly to academia and large corporations, but I'm keeping tabs on it. Today, I'm using ChatGPT to polish this post. I find generative AI fascinating and have started integrating Language Models (LLMs) into my work. I use GitHub Copilot for coding and have begun incorporating LLM tools into my applications. The open-source community is building remarkable tools based on publicly released model weights. I've been running a version of gpt4all on my Mac as a proof-of-concept service, alongside LocalAI for chat and speech-to-text functionalities.
If you haven't begun incorporating LLMs into your workflow or tech stack, you're falling behind. As the saying goes, "AI won't replace programmers; programmers using AI will replace programmers." Don't get left behind.