Original Article by Nicholas Collignon [here](https://open.substack.com/pub/ofmindsandcities/p/everyone-is-talking-about-reasoning?r=138la&utm_campaign=post&utm_medium=web&showWelcomeOnShare=false) Ref: [[The Misleading Allure of Anthropomorphizing AI]] | [[Probability x Uncertainty]] ### Background Picture a bustling city where every street corner brims with diverse minds debating new ways to move through traffic. Now swap that cityscape for the collective consciousness of researchers, developers, and curious onlookers all pondering how Artificial Intelligence—and the humans behind it—actually reason. We’ve never had so many brilliant people simultaneously scratching their heads over what it means to think, plan, and solve problems. From formal logic to creative leaps, reasoning has become the hottest neighborhood in the AI metropolis. #### Key Points 1. **Historically Rooted in Human Curiosity** As far back as Pascal and Bernoulli, scholars have grappled with how we interpret luck, evidence, and risk. They watched gamblers, judges, and everyday folks making decisions and turned those observations into probability theory. Those early frameworks still form the bedrock of AI research today. 2. **Formal Logic Meets Real-World Creativity** Pioneers like George Boole and Alan Turing used human problem-solving habits as models for systematic computation. Meanwhile, modern critics such as Subbarao Kambhampati emphasize “sound reasoning patterns” and formal correctness. But in open-ended domains—where messy human insights often spark breakthroughs—strict logical proofs can only go so far. 3. **Inference-Time Scaling and Emerging Capabilities** Up until late 2024, language models struggled to plan or handle multi-step tasks like the classic “Blocks World” puzzle. Now, new methods called “inference-time scaling” let them break down problems step by step, leading to noticeable leaps in performance. Still, the tug-of-war continues between those wanting cast-iron guarantees and those who see value in an occasionally fuzzy, human-like exploration of ideas. #### **So What?** For anyone building or using AI, the big takeaway is that true progress likely lies in blending the best of both worlds. We can’t ignore the rigor of formal proofs, especially in high-stakes applications that demand safety and accountability. At the same time, we shouldn’t dismiss the power of human-inspired reasoning: creative leaps, heuristics, and even our foibles can spark solutions where pure logic stalls. If you’re exploring AI—whether as a researcher, entrepreneur, or curious mind—recognize that new doors open when we combine methodical rules with the spontaneous street smarts of human thought. It’s like designing a city: by mixing careful planning with vibrant, messy life, we build something that truly thrives. And maybe, just maybe, we catch a glimpse of how minds—both human and machine—grow and evolve.