Startling AI results.

AI is not infallible. Whether it's the free or paid version. The output can vary greatly depending on the subject matter. Whether that's writing on a certain topic, doing research for a specific project, or asking it to do the heavy lifting on research data. You have be knowledgeable enough to spot bad output. I've had to correct AI output on building code-related issues. When I ask it to compare hi-fi component specs and performance in side-by-side comparisons, it will give me incorrect or outdated spec information that I have to spot. Does AI have useful applications? Yes, but if you're only using it for basic Q&A, or as a beefed-up search engine, then it will "feel" limited in its application. 

The trouble with using AI for basic queries is that it defaults to scraping relevant data it thinks you are requesting.  The last thing I would use AI for is health-related questions. Because, as others have mentioned, it doesn't know your specific health history, lifestyle, etc., that your regular PCP would know. AI also hallucinates a lot. 

Another thing I've experienced teaching grad courses is that students relying on AI output as 100% correct will not double-check output validity. In this respect, AI use can encourage intellectual laziness on the part of the user if they don't have enough experience knowing how to research for relevant information that isn't easily available. Especially behind pay-walled technical / academic sites. 

I asked Grok what the weakest link is and it said that I was.  Be careful how you phrase the question when speaking with our AI overlords. 🤣

@goldenways  "...I’m not sure what percentage of folks are using free version of AI vs paid services.  There is a world of difference between free and paid.   Exponentially. "

This was an interesting comment. I had always assumed it was simply  the size of the query or the number of queries that made the difference between paid and non-paid. I have been paid for over a year and use it all the time. 

So I posed this comment to ChatGPT. This is a summary of what it said. I definitely agree that for casual questions you will not see too much difference.

A more accurate version would be:

• There is often a significant difference between free and paid AI plans.
• The difference is not always exponential, and it varies by provider and by what you're trying to do.
• For casual questions, many users may see little difference.
• For complex research, writing, coding, analysis, long conversations, and large documents, paid plans can be dramatically better.

AI is the most powerful tool humans have invented for learning. But as with any tool you must learn how to use it. What its strengths are, limitations what to give it as inputs and you have to verify outputs when it is important. Effective use of the tool requires experience. Know how to craft prompts to get what you want and what to verify. I suppose this will be taught at some point. It is too new. 

Recently, an AI Expert described current AI systems as the equivalent of a stoned college teaching assistant.  Something worth keeping in mind when using AI.