Tonight, I had 5 minutes to learn w AI.

Usually, I will go to Claude for my 5 min w AI sessions, but tonight, I opted to try something different and checkout ChatGPT’s latest updates.

I entered my prompt into ChatGPT, setting the activity’s parameters.

ChatGPT returned:

The activity focused on AI learning, recognizing patterns, making decisions, etc.

The example ChatGPT gave related to a toddler learning to recognize a dog.

ChatGPT explained that it learns in a similar manner, through training data (& reinforced learning.)

The reply continued with an quiz that asked me to train an AI model to detect spam.

The first example was pretty clear.

The second one wasn’t quite as black and white, but I considered the language my bank would use in their notifications. They would never use the word “reminder,” so to me this was Spam.

It was a quick 5 minutes, so I opted for another activity.

This is where it got interesting.

So in this instance, the AI was trained to screen job applicants.

It was trained on past hiring data. In the past the company had hired more men than woman. The AI latched on to that and began prioritizing male applicants over equally qualified female candidates.

ChatGPT continued with a new scenario, asking me to spot the bias.

This one was a little more complicated. The AI was trying to predict which students would succeed in engineering.

The problem here: the AI was trained on a dataset which correlated success in engineering to attending a private school. So the AI develops a bias and begins favoring students that attended private schools.

Is the AI wrong? Are private school applicants be more likely to succeed?

Maybe. And also, maybe.

Yes, students that attended private schools may have had access to a more enriching education in preparation for university, than those that were not private schooled. But, that is NOT guaranteed.

Further, even if it were true, that doesn’t mean that students that attended a public, or other form of schooling, couldn’t also succeed.

Accordingly, everyone that meets the basic requirements for enrollment deserves a fair shot. And they must be evaluated on other things.

As I thought about, I generally don’t consider biases in the AI I use. But, I guess at this point the AI I am using is very ethically sound.

However, this may not always be the case.

This is something I need to keep in mind as I proceed forward in AI:

bias can exist.

Great exercise!