The Virtual Buddha: Learning How to Navigate the World

I’ve learned over my 50+ years on this blue marble the importance of being adaptable. These days, I’m adapting to perhaps one of the most transformative technology of our time: Artificial Intelligence (AI). My recent experiments with AI have been eye-opening, challenging some of my preconceptions and exposing to me both the real potential and limitations AI has.

Understanding AI: More than Just a Buzzword

When we talk about AI, we’re not just discussing a single monolithic technology but a diverse ecosystem of intelligent systems. Here are the major ones that most of are likely to interact with:

• Large Language Model (LLM): these are more conversational AI types and are probably the most well recognized globally today. Often these are trained on huge text datasets, able to understand and generate human-like response across a variety of topics.
• Retrieval-Augmented Generation (RAG): A more precise approach that combines language models with specific external data sources. The goal? Reducing the notorious “hallucinations” that can plague AI responses.
• Reinforcement Models: Think of AI that learns by interaction – like MIT’s robotic dog that improves its behaviour through experience.
• Multi-Modal Models: These go beyond text, integrating pictures, audio, video, and other forms of input to create more comprehensive understanding.
• Specialized Task Models: Narrow-focused AIs designed for specific domains, such as medical diagnosis assistance.

A quick note on hallucinations

Hallucinations are errors generated by AI based on knowledge it has. Often these are due to a lack of context and/or experience. This is often the biggest difference between the human experience and AI knowledge.

My Personal AI Experiment: Resumes and Revelations

When I found myself job hunting, I decided to experiment with AI as a writing assistant. My goal was simple: improve my resume and cover letters. What I discovered was far more nuanced than I expected.

Since it had been so long since I had last done a job search, I thought I’d enable some STMs (Specialized Task Models) around resume, cover letter and LinkedIn improvements. I used two major models found at JobScan and LazyPanda. The amount of hallucinations was nearly annoying. While JobScan faired better and was able to help me organize my resume so that it can be easily processed by recruiters that use the ATS model, either it added too much; removed too much; or diluted my knowledge.

Because of that, I turned to the more well-known public AIs of ChatGPT and Claude.AI. These presented less hallucinations for my resume and cover letters. Now whether it’ll be enough to get me hired, only time will tell.

Battle of the Bots: ChatGPT vs Claude.AI

Curious about their capabilities, I did ask both AIs to compare themselves. Their responses were telling:

Claude.AI’s Reponse

- I excel at nuanced reasoning, following complex instructions, and maintaining context

- ChatGPT is strong in creative tasks and broad knowledge generation

- The "best" AI depends on specific use cases and individual needs

ChatGPT’s Response:

It depends on what you're looking for! Both I (ChatGPT) and Claude.ai are advanced AI systems designed to assist users, but we have different strengths and approaches. Here's a general comparison:

Strengths of ChatGPT (OpenAI):

Language and Tone Versatility:
I excel at adapting tone and style to match various contexts, from professional to casual or creative writing.

Fine-Tuned Expertise:
My training incorporates diverse and extensive data, so I’m well-suited for problem-solving, brainstorming, and in-depth technical or academic discussions.

Tool Integrations:
In this environment, I can use tools for browsing, generating images, and even performing code execution, which makes me highly versatile for complex or dynamic tasks.

Stable and Predictable:
My responses tend to be consistent, logical, and grounded in clear reasoning.

Strengths of Claude.ai (Anthropic):

Longer Context Windows:
Claude is designed to handle longer inputs effectively, which can be helpful for summarizing or analyzing large documents.

Human-like Conversations:
It often feels very conversational and intuitive, making it appealing for natural interactions and brainstorming.

Ethical Alignment:
Built with a strong emphasis on safe and ethical interactions, Claude prioritizes user-centric engagement.

Choosing the Right AI for Your Needs:

For creative projects or tasks that require flexibility and a tailored tone: I may be a better fit.

For analyzing or working with very large text inputs: Claude’s longer memory might come in handy.

For general conversational tasks: It’s worth trying both to see which feels more natural to you.

Ultimately, both are excellent tools, and preferences often come down to personal experience and the specific task.

 Interestingly, while ChatGPT provided a more comprehensive breakdown, Claude felt more like a conversation and often, it was a very direct conversation. For my work on my resume, Claude’s ability to handle longer inputs and providing more concise feedback proved invaluable. At this point, I actually prefer Claude over ChatGPT for brainstorming.

Practical Tips for Leveraging AI

Through my experiments, I was able to develop a more practical approach when it comes to using AI effectively:

• Be specific and clear: Generic questions will get your generic answers. When I was trying to improve my resume, I didn’t just ask for general improvements. Instead, I specified: “Based on my uploaded resume how can I improve it for a position where the job has me creating technical curriculum?”
• Ask and ask again: just because the AI comes back with an answer doesn’t mean that answer is the final one. You can get it to revise or provide more guidance, again and again and again (within the limitations of a free account will allow)
• All AI have capabilities and limitations: being aware of those limitations is important. Always check the answer for validity. Certainly things have gotten better over the last couple of years but any AI can still hallucinate from time-to-time.
• It’s about collaboration: although all the FUD (Fear-Uncertainty and Doubt) suggests that AI will take over all our jobs, it likely won’t. Our jobs will change and it will become a more supportive tool for us to use. A recent article by Harvard Business Review seems to support this.
• Give feedback: tell the AI when the answer doesn’t match, is missing details or isn’t enough. AI learns from this feedback so it’s important to let it know if it’s on the right track or not.
• Ethical Use and privacy: Much like most of the internet, do not use anything that you wouldn’t share otherwise. My resume, for example, is out on LinkedIn as well so all that info is accessible. While I’d love to believe that AI is only used by humans for good, sometimes people will use it otherwise. AI is a tool and humans are the ones that will sway it towards evil.
  
  Microsoft’s early attempts at a public AI was in the form of an Twitter bot that rapidly became a “racist jerk“. It wasn’t the bot that was a problem; it was people.

Despite fears of AI replacing human jobs, the reality is more nuanced. As the HBR article suggested (referenced above), AI is more likely to transform our work, becoming a supportive tool that enhances what humans can do.

AI isn’t about creating a technological overlord. It’s about developing intelligent systems that can help us process information, solve problems, and unlock new possibilities. The key is approaching it with curiosity, critical thinking, and an open mind.

My journey with AI is just beginning, and each interaction teaches me something new. As technologies evolve, so must our understanding and approach. So have you explored AI tools in your professional or personal life? I’d love to hear about your experiences.

**Note: The image on this page was generated by ChatGPT and I had Claude help me organize things better.