Prompt and Result for "My Take on LLMs"

Prompt and Result for "My Take on LLMs"

Thanh Nguyen

Prompt

  • AI, in its simplest sense, is a "next word predictor"
  • I was skeptical at first
  • Started dabbling into using AI more with repetitive code
  • Find that it is useful if the code is simple and can be derived from something existed
  • Experimented with AI-aided reading; it felt better than "traditional" reading, too, as the progress is more interactive, and AI is able to refer to other sources following deeper explanation and interrogation
  • Experimented using AI as the first draft writer; it was also good as code base kick-starter (code structure)
  • From personal experience, AI's biggest limitation is context length
  • Example: for reading, if the book is too long then it cannot fit in, and we have to "chop" it to smaller parts
  • Nothing to back this argument (did not read a lot about it), but context length might hinder more complex problem solving as well
  • Apart from context length limitation, there is tacit knowledge that cannot be capture easily by text
  • Example: creating new and nuanced and complex code, apart from the code bases size, niche domain knowledge is not widely available
  • Example: writing a blog post without detailed outline will give surface level and non-cohesive writing
  • Useful tools: Claude for everything except coding; Supermaven for code completion
  • Final takeaway: despite the limitations, AI still is an useful tool; look forward to future advancements

Result

My Journey with AI: From Skeptic to Advocate
In its simplest form, AI is essentially a "next word predictor." This realization was my starting point on a journey that transformed me from a skeptic to an advocate for AI technology. Let me share my experience and insights with you.
Initial Skepticism
I'll admit, I was skeptical at first. The hype surrounding AI seemed overblown, and I couldn't see how a "next word predictor" could revolutionize our work and lives. However, curiosity got the better of me, and I decided to dip my toes into the AI waters.
Discovering AI's Potential in Coding
My first serious foray into AI was with repetitive coding tasks. I found that for simple code that could be derived from existing patterns, AI was surprisingly useful. It saved time and reduced errors in routine coding tasks, slowly chipping away at my skepticism.
AI-Aided Reading: A Game-Changer
Experimenting with AI-aided reading was a revelation. The experience felt superior to traditional reading in many ways:
The progress was more interactive.
AI could refer to other sources for deeper explanations.
It allowed for real-time interrogation of the text.
This approach made complex texts more accessible and engaging.
AI as a First Draft Writer and Code Structure Generator
Taking it a step further, I began using AI as a first draft writer for documents and as a kick-starter for code structure. This approach provided a solid foundation to build upon, significantly speeding up the initial phases of projects.
The Limitations: Context Length and Tacit Knowledge
Despite its benefits, AI isn't without limitations. From my personal experience, the most significant constraint is context length. For instance:
When reading long books, the entire text often can't fit into the AI's context window, requiring us to divide it into smaller parts.
This limitation might also hinder more complex problem-solving scenarios, although I haven't extensively researched this aspect.
Another crucial limitation is tacit knowledge - the kind of understanding that's difficult to capture in text form. This becomes evident when:
Creating new, nuanced, and complex code, especially in niche domains where knowledge isn't widely available.
Writing detailed blog posts without a comprehensive outline, which often results in surface-level, non-cohesive content.
Tools of the Trade
In my AI toolkit, I've found these tools particularly useful:
Claude: Excellent for a wide range of tasks, except coding.
Supermaven: My go-to for code completion.
The Takeaway
Despite its current limitations, AI has proven to be an invaluable tool in my workflow. It has transformed how I approach various tasks, from coding to writing and research. As we look to the future, I'm excited about the potential advancements in AI technology. Who knows what new doors these developments will open?
Remember, AI is a tool to augment human intelligence, not replace it. By understanding its strengths and limitations, we can harness its power to enhance our productivity and creativity. The key is to approach AI with an open mind and a willingness to experiment.
What has been your experience with AI? I'd love to hear your thoughts and experiences in the comments below!

Report Page