👽 AI - The Pitfalls...

[webtrekker]

AI, and ChatGPT in particular, may seem very clever at giving almost instant answers to all kinds of queries, but these answers are not always reliable!

Here's an example of what I mean ...

 

In ChatGPT, and AI in general, a LOT depends on the way the Prompt is formed.

 

For instance, even though ChatGPT is a LLM model (Large Language Model), it is still capable of straightforward mathematical problems, GIVEN THE RIGHT PROMPT!

 

 

Example:

 

Prompt:

Calculate SQRT(78.9)

 

ChatGPT answered:

The square root of 78.9 is approximately:

√78.9≈8.878

 

**THIS IS THE WRONG ANSWER!**

 

=====================================================

 

Prompt:
You are an expert in Mathematics. Calculate SQRT(78.9)

 

ChatGPT answered:

Certainly! The square root of 78.9 is calculated as follows:

√78.9≈8.882

This value is precise up to three decimal places. Let me know if you need further calculations or explanations!

 

**THIS IS THE CORRECT ANSWER!**

 

=====================================================

 

Note the difference between the Prompts. In the second Prompt, I gave ChatGPT a role to play. I told it that it was an expert in Mathematics. This makes a big difference to the way it behaves, and it does, in fact, end up with the correct answer.

 

You need to be very tactful in preparing Prompts for AI, and always check the results wherever possible.

 

Edited January 26 by webtrekker

[Grumpy Grapes]

Human AI Trainers are being used to make AI more effective and reliable, so we're told.

 

"Introduction to AI Trainer Jobs

"An AI Trainer plays a pivotal role in the development and refinement of AI systems. They are responsible for teaching AI models how to interpret and process data accurately. This involves a combination of data annotation, algorithm training, and continuous improvement of AI behaviours. As AI becomes increasingly integrated into various sectors, the demand for skilled AI trainers in the UK is growing exponentially.

 

"Why AI Trainers are Essential

"AI systems rely heavily on high-quality data and proper training to function effectively. Without the guidance of AI trainers:

 

"Misinterpretations Increase: AI models may misunderstand data, leading to errors.

 

Inefficiencies Arise: Poorly trained AI systems can become inefficient, wasting resources.

 

"User Trust Diminishes: Inaccurate AI behaviours can erode user confidence.

 

"AI trainers ensure that AI systems are reliable, accurate, and aligned with human expectations, making them indispensable in the AI development process.

 

https://artificialintelligencejobs.co.uk/industry-insights/ai-trainer-jobs-a-comprehensive-guide-for-uk-job-seekers#:~:text=An AI Trainer plays a pivotal role in,algorithm training%2C and continuous improvement of AI behaviours.

 

 

[Grumpy Owl]

On 1/26/2025 at 1:57 AM, webtrekker said:

Note the difference between the Prompts. In the second Prompt, I gave ChatGPT a role to play. I told it that it was an expert in Mathematics. This makes a big difference to the way it behaves, and it does, in fact, end up with the correct answer.

 

You need to be very tactful in preparing Prompts for AI, and always check the results wherever possible.

 

I've been dabbling with AI at work, and yes, the 'prompt' is something I have learned about when it comes to getting the best out of ChatGPT (or any other similar AI model).

 

By default, their response is always 'very chatty'. But with a suitable prompt, you can almost instruct them to 'get to the point' and just give an answer, without any superfluous 'waffle'.