AI and the Challenges Ahead

 

I have a growing curiosity about Artificial Intelligence (AI) and I have been playing with the ChatGPT app to understand it’s capabilities and where we are in 2024!

First and foremost, you need to understand that this is a language-based app. It uses enormous computing power to interpret what I am asking and then incorporate all it has learned to provide a response.  It’s training ended in 2022 so current events, such as the Israeli – Hamas war are not in its realm of knowledge.

My first attempt to use ChatGPT was particularly eye opening.  I asked what I assumed to be a simple question…

Can you create a list of Presidents of the United States ordered by their age at inauguration?

The result was surprising.  As I am sure you know Joe Biden is the 46^th president in our nation’s history.  Surprisingly, ChatGPT produced a list of only 41 people!  I did not scrutinize the list, but it was apparent to me that Mr. Biden was not on the list (he should have been last being our oldest president!)

So, my follow up question was… Why is Joe Biden not on the list?

This solicited an apology from the app and a revised list, now with 42 presidents.

I looked a little closer and discovered that one of the presidents still missing was George Washington!

Next, I asked that the Inauguration Date be included.  This resulted in the app correctly identifying all 46 presidents.

I watched a Ted Talk by Yejin Choi entitled: “Why AI is incredibly smart and shockingly stupid?”  In it she described a scenario where she asked the app…

(I am paraphrasing)

If it takes 5 hours to fully dry 5 shirts on the line, how long will it take to dry 10 shirts?

The answer… 10 hours of course!

The system cleverly calculated that 5 shirts in 5 hours = 1 hour per shirt it then multiplied the result by 10 and got 10 hours!

Here is the link to her talk, it’s worth 15 minutes of your time

https://www.ted.com/talks/yejin_choi_why_ai_is_incredibly_smart_and_shockingly_stupid?referrer=playlist-the_must_watch_ted_talks_on_ai_from_2023&autoplay=true

To recreate this result on my own.  I decided to try a similar question.  I confess that I set it up which may not have been fair, but I started with, “Can I ask you a math question?”  Then I asked…

If it takes a bus with 4 people on it 4 hours to get from New York City to Washington DC, how long will it take to get 8 people from NYC to DC?

As you might expect, the answer involved proportionality…

“To solve this, we can use the concept of proportionality.  If the time it takes is directly proportional to the number of people, then we can set up a proportion:

4 hours / 4 people = x hours / 8 people

Now we can solve for x:

X = (4 hours * 8 people) / 4 people

X = 8 hours

So, it will take 8 hours for the bus with eight people to travel from New York City to Washington DC.”

Bring on the SAT’s!!

Finally, I tried the same question again.  This time not setting it up as a math problem. 

“Assuming all other factors remain constant, it would still take 4 hours for the bus with eight people to travel from New York City to Washington DC.  The number of people on the bus doesn’t affect the travel time, as long as the bus itself can accommodate them comfortably.”

Which of course is a MUCH better answer!

 When you think about the results here you can see that AI is far from perfect.  It knows a lot of stuff, its explanation of proportionality was clear and accurate, however, it lacks the basic common sense that my 7-year-old grandson can apply instantly.

AI is a tool, think of it as a hammer.  We will need to learn to use it so as not to leave big dents in the wall when we miss the nail!  Everything we get from AI needs to be examined carefully and validated before it is used.  Don’t rely on it to be right the first time.

In June of last year Attorney Steven A. Schwartz needed to file a brief in Federal District Court in Manhattan.  Instead of doing the necessary work to prepare the motion, Mr. Schwartz utilized AI to prepare it for him. 

After the brief was submitted, opposing council complained to the judge that they were unable to find ANY of the cases referenced in the document.  In fact, ChatGPT made up ALL of the supporting material!

Mr. Schwartz was excoriated by the judge and left the courtroom vert embarrassed!

AI can and will be a valuable tool.  It can help us to figure out complex problems and help us to be much more efficient.  However, in my opinion, it may never be a reliable substitute for common sense!