Generative AI began as little more than an advanced autocomplete. Today, with reasoning models like OpenAI's o3, machines can approach problems step by step, much like engineers do. This edition explores how that evolution is reshaping debugging, design, and everyday problem-solving.
Generative AI started out like an autocomplete on steroids. Now, with reasoning models such as OpenAI's o3, machines work step by step, much like engineers do. This installment looks at how that shift changes debugging, design, and everyday problem-solving.
OpenAI Platform
For a while now, generative AI has felt like a very clever -- but very impatient -- colleague. You ask a question, and it spits out the most statistically likely answer, often a list of shotgun-style guesses. It's impressive, sure, but it's a parlor trick, not genuine thought. The real story of late is not just about more data or layers. Rather, the market is working very hard on reasoning.
In the previous installment of this column, I made some basic reasoning comparisons between different models. The one I'm testing, OpenAI's o3, works differently. It doesn't just blurt out an answer, but rather takes its time, reasoning step-by-step, much like an engineer at a bench methodically checking voltages and pro...