"You're holding it wrong," Steve Jobs's famous response during the iPhone's Antennagate fiasco— Similarly, LLMs have been in the mainstream for almost two years now, and they have improved significantly with each iteration. However, when I ask people (and I do ask almost everyone), "How are you using generative AI in your daily lives?" I rarely hear of anyone leveraging these tools to their full potential or solving real-world problems.
The most frequent use case among my peers is basic text generation, like crafting job postings. Many who try these tools a few times find them underwhelming and abandon them. The likely reason? They're using them wrong.
How to Better Use LLMs
Change Your Mindset: Don't treat these tools like a Google search; they are fundamentally different. Consider the early days of television, where broadcasts were essentially just a camera pointed at a radio—people were watching a radio on TV! We often use new tools within the framework and mindset of old ones.
Choose Your Weapon: Before defaulting to ChatGPT or Gemini, determine which LLM is best suited to address your specific problems. Each LLM excels at different tasks in unique ways.
Craft Your Prompt: Take your time to explain your problem, provide context (LLMs thrive on context), give examples, and even ask the AI to explain its steps (a method known as Chain of Thought prompting, which significantly improves outputs).
Speaking of replacements, I've completely switched from searching the internet with Google to using an LLM called Perplexity. It's as if Google and ChatGPT had a baby. This tool not only delivers detailed, summarized answers that are grounded in reality, but it also cites its sources! I highly recommend giving Perplexity a try.
How I Am Using AI
I've found numerous ways to integrate these tools into my daily life and work. Here's how I'm leveraging AI:
Text Enhancement: For text enhancement, I often put what I write into an LLM to check for spelling, punctuation, grammar, and readability enhancements. I also ask LLMs for feedback and suggestions for titles. I don't have it write from scratch for me, but I'll use it to make sure I get my point across or tell me where I can improve.
Coding Assistance: I use GitHub Copilot, as well as Claude and ChatGPT to enhance and help me code. I even share terminal output with them. I also use some open source coding models which I will talk more about in another post.
Image Generation: I use Midjourney and DALL-E to generate images for various projects.
Learning Tool: I use AI to learn about a wide variety of subjects as well, expanding my knowledge base
efficiently. (But will often check multiple sources or LLMs to avoid learning from a hallucination.)
By integrating AI into these aspects of my work and life, I've found that I can be more productive, creative, and informed. I encourage everyone to experiment with these tools, find what works best for you, and always approach them with a curious and critical mind. The potential is huge, but realizing that potential requires us to adapt our thinking and use these tools in innovative ways.a
Comments