The Value of Humans - AI Errors

Is AI Overhyped?

Maybe. But this is a debate is nothing but a hole that will only dig deeper and deeper

Here's the reality:

Here’s the reality: Being human is more valuable than ever, and the skills and insights that our Humanity brings to the table are becoming increasingly apparent.

I’ve been using several paid AI tools to pull data from publicly available sources, and guess what? The mistakes are plenty and important.

So, what's the takeaway?

Understanding what AI can—and can’t—do is crucial. This way, we can use AI as a personal turbocharger to get more done and achieve better results, while sidestepping common pitfalls.

As I have shared repeatedly on webinars and conferences, AI simply does not have the ability to know what it does NOT know.

(image created with Dall-e)

Key Insights:

1. Data gathering errors:

When I asked ChatGPT and Claude AI (both paid versions) to gather detailed information about ICE facilities and their managing companies, mistakes occurred in 5 rounds of very clear and specific prompts. The error rate was around 72%. Mistakes included omitting data, including incorrect information, and worsening errors despite clarifications.

2. Web crawling limitations:

There are a number of web and data architectures that AI canNot crawl, which means that such specific tasks will require additional solutions and proofreading.

3. Coding is alive and well:

When NVIDIA CEO claimed that Coding was dead, he was just looking for attention and free publicity. Research and experience show that human involvement is still crucial in many fields, including coding. Be skeptical of dramatic claims from CEOs playing Fonzi—they often don’t have all the answers.

What does this mean for Us Mortals?

1. It means that the reality as we live in is shaped by us, humans, which means that it is riddled with complexities that require our experience and expertise to find solutions.

2. That when we put human insight and the right AI tools in the blender, we can create New Mother Sauces.

3. And, despite concerns about job losses, AI appears to be creating more opportunities than it eliminates.

In case you are curious:

I used ChatGPT(paid) and Claude AI (paid).

The request was about ICE facilities, and the companies that manage them (more specific data was requested), and I got mistakes in 5 rounds of hyper clear and detailed prompts.

What type of errors were made?

1. Omitted Data. Items that clearly meet the request were not included.

2. Incorrect Data. Items that do NOT meet the criteria detailed in the prompt were included.

3. Persistent errors. Mistakes continued despite corrections and clarifications.

I’m a student of AI, focusing on Human Communication, Language, and Culture. I help organizations develop workflows that maximize AI’s potential while avoiding common pitfalls.

Keywords:

#AIData #AIDatasearch #AIdoesNOTknow #AIErrors #ICE #Detentionfacilities