No Need to Fear the AI Coming Apocalypse—AI Still Can’t Write Code

Title: The AI Coding Dilemma: Why You Shouldn’t Fear the Rise of Intelligent Machines

In recent days, I’ve found myself immersed in an unexpected challenge with two prominent AI models: ChatGPT and CLAUDE. What began as a simple attempt to create some graphics for Substack quickly spiraled into an unexpected odyssey. My initial plan was to leverage PowerPoint for this task, but curiosity got the better of me. I thought, “Why not explore the potential of AI for coding?”

What I anticipated would be a straightforward shortcut turned into a perplexing adventure. While I believed I could save considerable time, reality quickly presented a different picture. Each time I attempted to resolve one coding issue, a slew of new problems would arise, creating an exhausting cycle reminiscent of a game of whack-a-mole. This experience was peppered with frustration, and I couldn’t help but refer to it as a “cluster…” well, you can imagine my choice of words.

After several disappointing attempts with the AI, I reverted to my original plan and crafted a polished PowerPoint presentation in just a little over three hours. This experience served as a reminder that while AI possesses impressive capabilities, it often mimics the learning curve of a novice programmer on their first assignment.

Certainly, advancements in AI will come from techniques like Reinforcement Learning from Human Feedback, which will help these systems become more adept. Experts will play a fundamental role in refining these automated coders, and improvements are undoubtedly on the horizon.

However, the prevailing fears surrounding AI capabilities, particularly in coding, are often overstated or fueled by those with an agenda. Instead of succumbing to anxiety about an impending AI takeover, perhaps it’s wise to take a step back. Pour a glass of Scotch, take a moment to relax, and remember that the technology, at least for now, is not quite as advanced as some would have you believe.