Using AI
Introduction
The introduction of generative AI tools like OpenAI's GPT models has begun to change how knowledge workers approach their jobs. As I've experimented with integrating these tools into my daily workflow over the last couple of years, these tools have reminded me of the ship's computer in the Star Trek TV shows and movies, particularly Star Trek: The Next Generation.
Describe the computer in Star Trek: The Next Generation
The computer in Star Trek: The Next Generation is an iconic example of a conversational AI: always available, context-aware, and able to answer complex queries or execute tasks with natural language. It doesn't make decisions for the crew, but instead provides information, runs simulations, and helps them reason through problems. In other words, the computer augments human expertise rather than replacing it.
The following clip from Star Trek: The Next Generation is from a longer video that shows the crew interacting with the computer in several episodes. The clip from 11:21 to 14:09 is from the episode “Identity Crisis” (Season 4, Episode 18), where Geordi La Forge consults the computer to analyze strange occurences during an away mission.
You may watch the entire video on YouTube.
What is LLM?
A Large Language Model (LLM) is an AI system trained on vast amounts of text data, enabling it to generate human-like responses, summarize information, and assist with a wide range of tasks using natural language.
(start ai) But that's the technical definition. From a practical standpoint, an LLM is like having a remarkably well-read programming partner who never gets tired, never gets frustrated, and is always willing to help you think through a problem—even when that problem is something as mundane as “Why won't this regex work?” or as complex as “How should I architect this distributed system?”
The key thing to understand about LLMs is that they're not databases of facts—they're pattern recognition engines trained on human language. This makes them incredibly powerful for tasks that involve understanding context, generating code, and explaining complex concepts, but it also means they can confidently give you incorrect information if you're not careful. (end ai)
Using LLMs Like the Star Trek Computer
Modern LLMs, such as GPT-based systems, are the closest we've come to the Star Trek computer. They can answer questions, generate code, and help solve problems conversationally, though they still have limitations compared to the fictional ideal.
(start ai) The parallel isn't perfect, though. The Star Trek computer has access to all of Starfleet's databases and sensors—it knows exactly what's happening on the ship at any given moment. Our LLMs, by contrast, only know what we tell them about our specific situation. This context limitation is both a weakness and, oddly, a strength. It forces us to be more deliberate about how we frame our problems. (end ai)
Geordi is the expert
In Star Trek, the crew relies on the computer for information, but domain experts like Geordi La Forge interpret and apply that information. Similarly, LLMs are tools to augment human expertise, not replace it.
(start ai) This is perhaps the most important lesson I've learned from working with AI tools: the human is still the expert. Geordi doesn't just ask the computer to fix the warp core and walk away. He uses the computer to run diagnostics, model different scenarios, and access technical specifications, but he applies his engineering knowledge to interpret the results and make decisions.
The same principle applies to coding with LLMs. I might ask an AI to generate a function, but I'm the one who understands the broader architecture, knows the business requirements, and can spot when the generated code doesn't quite fit the context. (end ai)
How I Use LLMs to Develop Software
LLMs have become an integral part of my software development workflow. I use them to:
- Generate code snippets and boilerplate
- Explain complex concepts or unfamiliar APIs
- Review and refactor code for clarity and efficiency
- Brainstorm solutions to tricky problems
- Summarize documentation or error messages
By treating LLMs as collaborative partners—much like the Star Trek crew interacts with their computer—I can work more efficiently and creatively.