Everyday Artificial Intelligence

 

Can artificial intelligence support everyday decisions?  In other words would our lives be better off if we ceded all our decision making to AI?  As I have said in past blog entries, as a retiree I am no longer a slave to corporate rules (yes a little dramatic).  Ninety percent of how I spend my time is up to me.  I make decisions as to what to wear each day.  Even getting out of bed is more of a personal decision.  For the past 40 years I got out of bed and showered because I had to go to work.  Now I can decide whether or not getting out of bed is a good idea.  So in reality if I do not want to get out of bed I do not have to.  However I am fairly certain that if I made a habit of spend all day in bed I would soon find myself on the street.  I doubt my wife would have much patience for a lazy good for nothing husband who wouldn't get off his rear end.

A little understanding of what AI is would help here.  The better we understand it the better we can assess how it can apply to our lives.  I actually used chat GPT to tell me what AI was. Here is my summary of what chat GPT returned when asked "How does artificial intelligence work?".  The process begins with data.  The first step is to collect the data and the second is to preprocess it.  Data is collected because it is needed to make decisions.  Preprocessing the data makes it easier to analyze, and search.  If you think of the old style physical phone books as the data source.  The data in it is structured to allow the user to find a phone number for a given name.  If you have a phone number and needed to find the person associated with it a standard phone book would be quite useless.  Using a standard phone to find the name associated with a given phone number would require one to begin with the first entry and go line by line through the phone book.  Every phone number would need to be compared to the given number for a match.  Once the numbers were matched the name would then be known.  For a normal city the standard phone book may contain a couple hundred thousand entries.  It is easy to speculate that searching for a specific number would take a considerable amount of time.

Data is also placed in sets or "Buckets" to make searching easier.  I asked chatgpt for the best road hockey stick.  To reply to this AI would be looking at data that resides in the road hockey stick "bucket".  There would be no need to search through data on all sporting goods equipment.  It would be like any one of us going to a sporting goods store to purchase a road hockey stick.  We would not spend time in every area of the store looking for road hockey sticks.  We would not go to the baseball section pick up a bat and determine that it would make a poor road hockey stick and return it to the rack.  We would go to a store layout map and determine the road hockey area and proceed there.  Once there we would begin to analyze the road hockey stick offerings to find the best one.

The goal of all this preprocessing is to make the data easy to analyze.  This allows AI to return answers quickly.  When replies are immediate and concise it appears as if you are chatting with an "expert' in a particular field and not a computer.  How ever this has a down side.  It takes time to preprocess and verify all this data.  I asked chatgpt "who has won the most best male actor in a leading role oscars?"  the reply came back with a time frame limitation.  The reply stated that as of 2022 Daniel Day Lewis had won 3 best actor in a leading role Oscar's.  The data set that chat gpt was using was several months out of date.  Decisions based on out of date data can be risky.

Step 3 involves selecting an algorithm.  Selection is determined by the type of data.  I am comfortable with steps one and two (data collection and preprocessing).  Step 3 puts me in a state of considerable unease.  I am at a loss of how to adequately explain these algorithms.  The best I can do is explore the responses to two related but different queries.  The first is one from above;  "who has won the most best male actor in a leading role oscars?".  The data for this question is very definitive.  It would be a list of all the winners and the response would be computed by adding up each winners total wins and presenting the actor with the most wins (or list if it is a tie).  I then asked chatgpt "who is the best male actor?"  As new data is updated the algorithm would update the output.  If Tom Hanks wins 2 more Oscars (giving him 4 wins) the response would update.

This second query is very subjective, where as the first is very objective.  The algorithm for the second question would need to be quite different than the first.  The second question would have to scan multiple data sources.  It may have pulled a list of actors and then scanned reviews to pare down the list.  Chatgpt returned the following actors: Marlon Brando, Robert De Niro, Daniel Day Lewis, Al Pacino, Denzel Washington, Tom Hanks and Leonardo DiCaprio.  The first thing that struck me was that all of these have won at least one Best actor Oscar.  Maybe the algorithm began with the list of best actor winners and then pared it down from there.  That could be the reason for Tom Cruise not being in the list in spite of his success.  Other omissions; Harrison Ford, Edward Norton, Samuel L Jackson, just to name a few.

I decided to test out the algorithm by asking:

"if Tom Hanks wins 2 more best male actor in a leading role would he have the most wins?"

Yes, if Tom Hanks were to win two more Oscars for Best Actor in a Leading Role, he would indeed tie Daniel Day-Lewis for the most wins in that category. As of my last update in January 2022, Daniel Day-Lewis holds the record for the most wins in the Best Actor category, with three wins for his performances in "My Left Foot" (1989), "There Will Be Blood" (2007), and "Lincoln" (2012).

If Tom Hanks were to win two more Best Actor Oscars, he would also have three wins in the category, placing him on par with Daniel Day-Lewis. Tom Hanks has previously won the Oscar for Best Actor twice, for his roles in "Philadelphia" (1993) and "Forrest Gump" (1994). Therefore, two additional wins would indeed tie him for the record.

Apparently Chatgpt cannot add (2 + 2 = 4)

Based on the time limitations and the fact that it cannot add correctly I will be making my own decisions.

One last Chatgpt response:

Me: I do not have a dress code to worry about should I wear pants today?

Chatgpt: If you don't have any specific dress code to adhere to and the weather permits, whether or not to wear pants ultimately comes down to your personal preference and comfort. If you feel more comfortable and confident wearing pants, then go ahead and wear them. However, if you prefer the freedom and breathability of wearing shorts, that's perfectly fine too. Consider factors such as the temperature, your planned activities for the day, and what you feel most comfortable in. Ultimately, the choice is yours!

Yep the choice is mine!