Sign in Subscribe

Topic

inside-the-machine

Exploring how AI and generative systems work through visual metaphors and accessible explanations. Friday posts demystifying complex technologies.

Data Stacking Magic: How Tensors Help AI See What We Can't

We can picture a list of numbers or even a table of data. But what happens when information needs 5, 10, or 100 "directions" at once? Tensors let AI stack data in ways our brains simply can't visualize. Is this why AI sometimes makes connections that completely surprise us humans?

Words as Math: The Shape of Meaning

Type "king - man + woman" into an AI and you get "queen." This isn't magic—it's math. When AI turns language into numbers, meaning isn't in definitions but in patterns. Could this be how meaning really works in our minds too—not as dictionary entries, but as a map of connections?

How (LLM) AI Sees Language Through Token-Colored Glasses

Ever wonder why AI misunderstands simple concepts yet writes perfect poetry? It never sees words—just statistical puzzle pieces called tokens. Your "cat" becomes numerical patterns with no inherent catness. What happens when thinking is built on language fragments?

The Conversational Advantage (To LLM Prompting)

Is AI actually smarter when we stop treating it like a computer? When we just talk—messy, tangential, and human—AI systems access neural pathways they can't reach through rigid commands. Natural conversation might unlocks creative connections that structured prompts can't.

Beyond Shiny Objects: When Technology Finally Matters (to me)

I spent years watching people worship at the altar of shiny gadgets, puzzled by their devotion. Then AI showed up – not as another toy, but as a mirror reflecting my thoughts in ways I'd never seen before. The magic wasn't in what it did, but in how it transformed my thinking itself.

1

Content Generation

Articles are composed using Claude 3.7 Sonnet (developed by Anthropic) with customized context documents and specialized instructions. The creation process begins with my initial concepts and topics, followed by an iterative dialogue where I refine ideas and request additional exploration. This collaborative approach combines my unique perspectives with AI's ability to help articulate complex ideas, resulting in content that remains authentically mine while benefiting from enhanced expression and organization.

2

Research Methodology

Factual information presented in these articles is supported by citations to original sources. A significant portion (over 50%) of these referenced materials were identified through research queries processed by Perplexity.ai using a customized research and source validation methodology. This approach enables more comprehensive research by expanding the range of sources I can discover and incorporate, while maintaining rigorous standards for accuracy and relevance through my careful review of all cited materials.

3

Editorial Oversight & Accuracy

While the final text is generated by AI tools, all content has been reviewed, approved, and published under my editorial direction. While efforts have been made to ensure accuracy, I cannot guarantee that all AI-generated content is error-free. Readers should verify critical information from primary sources before relying on it. The views expressed are my own and do not necessarily reflect those of the AI tools used. This disclosure is provided in the interest of transparency and ethical AI use.