A Perspective After 40 Years in Technology
AI isn’t as new as the headlines suggest. While it feels like it arrived overnight, the foundations have been around for decades, mostly in specialized labs for testing and experimentation. We’ve actually been using early forms of AI for years; think of “Hey Siri” or the grammar checkers that suggest better ways to word your sentences.
It’s Not a Better Search Engine—It’s a Different Engine Entirely
Many people think of AI as a smarter version of Google, but this is a misconception. When searching online, you receive an index of links to websites. Using AI involves “Prompt Engineering,” which means crafting a detailed request so the AI can synthesize real data into specific, cohesive answers rather than a list of sites.
Meet the Major Players
There are many types of AI; some make music, others create videos, but the most popular today are Large Language Models (LLMs). You’ve likely heard these names:
- ChatGPT: The “all-rounder” and the first to make a massive public splash.
- Gemini (Google): Leverages the massive Google ecosystem and real-time information.
- Copilot (Microsoft): Ideal for office tasks; it’s built directly into tools like Excel and Word.
- Claude: Known for having a more “human” feel, often favored by writers.
- DeepSeek: A recent entry that gained attention for performing complex computing using less hardware power (GPUs).
But with so many AI options available, a common question is: why do their responses vary?
If you give the same prompt, for example, asking for a summary of Civil War battles and their impact on Northern strategy to three different AIs, you will get three different responses.
This isn’t necessarily because one is “wrong.” It’s because they are built differently. Each model was trained on different data sets (books, articles, videos) and uses different “learning methods.” Think of it like asking three different historians the same question; their perspectives change based on the books they’ve read and their specific area of expertise.
A good prompt
The “Deep Dive” Analysis Prompt
Act as a military historian. Analyze the Battle of Gettysburg (July 1863) and provide a concise summary covering:
- The Tactical Turning Point: Identify the specific moment the momentum shifted.
- Logistical Impact: How did this battle affect the South’s ability to supply its troops for the remainder of the year?
- Strategic Consequence: Explain how this victory changed the North’s overarching ‘Anaconda Plan’ strategy.
Please provide the answer in a structured format suitable for a quick briefing.”
The Digital Blueprint: The Transformer
Modern AI uses a specific architecture called a Transformer. Its secret sauce is a mechanism called Self-Attention. When you give the AI a sentence, it doesn’t just read left-to-right; it looks at every word simultaneously. It assigns “weights” to words, essentially deciding which words are the most important to the meaning of the others. This allows it to understand that in the phrase “The bank of the river,” the word “bank” refers to land, not a financial institution.
The Engine: Neural Networks & Layers
An LLM is made of hundreds of these layers stacked on top of each other.
- Input Layer: Breaks your text into “tokens” (chunks of characters).
- Hidden Layers: This is where the actual “thinking” happens. As data passes through, it’s multiplied by billions of tiny numbers called parameters. These parameters are the model’s “experience”—they determine how the signal changes as it moves through the web.
- Output Layer: Converts those processed numbers back into the word it predicts should come next.
The Muscle: The GPU
While a standard computer processor (CPU) is like a world-class chef who can do one complex thing at a time, a GPU (Graphics Processing Unit) is like a thousand line cooks working in parallel.
Because an LLM has to perform billions of simple math calculations (multiplication and addition) at the exact same time to process a single sentence, the CPU would be too slow. The GPU’s ability to handle parallel processing enabled AI to move from a slow lab experiment to a real-time tool you can chat with on your phone.
The Takeaway
AI is a tool for synthesis. Instead of manually clicking through numerous links to find an answer, the AI examines those sources and provides the summary you need. The value is not in reaching a specific destination, but in generating relevant results efficiently.
The next part of the series will address the impact of AI: jobs, the number of players before some fall away, and the potential long-term impacts on society.
Leave a Reply to kwholley63Cancel reply