Ovi -
we cover every issue
newsletterNewsletter
subscribeSubscribe
contactContact
searchSearch
worldwide creative inspiration  
Ovi Bookshop - Free Ebook
Ovi Greece
Ovi Language
Michael R. Czinkota: As I See It...
Stop violence against women
Tony Zuvela - Cartoons, Illustrations
International Red Cross and Red Crescent Movement
 
BBC News :   - 
iBite :   - 
GermanGreekEnglishSpanishFinnishFrenchItalianPortugueseSwedish
Artificial intelligence and computer-assisted decision-making Artificial intelligence and computer-assisted decision-making
by Joseph Gatt
2020-12-19 11:19:59
Print - Comment - Send to a Friend - More from this Author
DeliciousRedditFacebookDigg! StumbleUpon

OK. AI can defeat the very best at chess of at Chinese checkers. Those are two highly strategic games, where computers are programmed in ways not to attack, but to be very, very defensive.

That is, in chess or checkers, computers know how to protect their pawns or checkers from being “eaten” by way of complex mathematical calculations. So the computer wears out the chess or checkers human player by playing very defensive games. The human player gets exhausted by not being able to “eat” the computer's pawns, and the computer wins, or more often, it's a draw.

But, AI probably will never be able to predict the stock market. Nor can AI predict sales of a product. Nor can AI take rational decision-making.

arint001_400Here's what AI knows. AI knows that big brand soft drink stocks and shares tend to shoot up in the summer and fall in the winter. AI knows that oil and gas company stocks tend to shoot up around November, and fall starting from the end of March. AI knows that any big brand manufacturing coats is going to have good share value in October or November. AI knows that cruise ship and resort share values tend to shoot up around May and June, and fall mid-August.

Because those have been the trends for years. Most people purchase homes around the month of September. Most people vacation in the summer. Most people visit their families around the Christmas holidays. And most people purchase furniture and stationary around late August to early September. And most people purchase toys and electronics around late November to mid-December.

AI can also identify a few patterns. Most couples get engaged around the months of March to May, because those are the months where the couple will tend to identify stability in their professional life. The work year tends to start in September, and if you're thriving at your job by March, you know you have a good gig going. March to May is also movie theater season, and that's when most couples go to the movies rather frequently.

AI can come up with complicated algorithms and calculations to determine how weather factors, environmental factors and economic indicators in your area will translate to an estimated customer base visit to your store. Those are estimations, and they tend to be rather accurate.

AI can also identify consumer trends by analyzing key words typed in search engines. AI can identify consumer trends based on friendships and affiliations on social media. AI can even predict or estimate that crime will be committed in certain areas, based on factors such as online behavior and crime trends.

But, the question is, what do we humans do with all this information? The information should not and does not take decisions for us. The information assists us in our decision-making, and we humans should end up deciding how to act based on the information, along with other sources.

Just like AI can defeat the very best at chess or Chinese checkers, AI robots could also defeat the very best at soccer or basketball, running faster, passing more accurately and shooting more accurately than Pelé or Michael Jordan ever did.

The problem with AI is that some past trends evolve and look nothing like the future trends. For example, coats indeed sell very well in October and November, but the kind of coat that sells really well varies a great deal from year to year. Some years it is sports coats, other years it is more stylish coats. Same goes for shorts or shoes. The brand that comes out as a winner varies greatly from year to year.

Other trend that AI can't predict: that impulsive moment that leads an individual or a company to near collapse. Could be a surprise resignation by the CEO. Could be an unexpected scandal. Could be an unexpected lawsuit. Could be an impulsive legal change that changes the game completely. Could be a surprise general strike that leads to collective resignations.

Just like AI can predict hurricanes, more often than not, it can not predict the intensity of hurricanes. Sometimes a category 5 hurricane is announced and you end up with a category 1 hurricane. In some cases, a category 3 hurricane is announced, and it ends up being a category 5 hurricanes.

AI is a lot worse than that at making predictions. Just to give an example, AI models had predicted that because more people were bachelors living in bachelor pads, craft food sales would explode. Problem is: bachelors weren't purchasing craft food like macaroni and cheese, they were having their meals delivered home by local restaurants, or they were eating out.

Other example: AI models predicted that social media users would be more willing to share information. Problem is, social media made users compete against each other and made users more hesitant to share information. So social media users were less informed than in times where all the information was on the bulletin board.

Yet another example before I stop. Elections. Before the days of AI, when polls were done by calling representative samples on the phone, polls were more or less reliable, and there were some notable mistakes. The 2002 French presidential election is a good example, where Socialist candidate Jospin was comfortably in second place according to the polls. Some analysts did ring warning bells, by saying that ultra-nationalist candidate Le Pen could well be in second place, but those warnings were ignored. In the end Le Pen did finish second. That is because those polled on the phone tended not to confess that they were going to vote for Le Pen.  

But in the age of AI, the algorithms look at data. And the problem is that the data that users provide is often not very reliable. Internet users often provide data on their aspirations and who they aspire to be rather than data on who they actually are.

For example, most Internet users are probably not going to Google racist information or read racist pamphlets or racist blogs. But that doesn't mean they don't have racist ideas. Or, Internet users could leave no trace of socialism in their Internet activities, they might even behave in ways a typical hardened capitalist would. And yet, when they watch TV or read the papers, they become convinced socialists.

Here's a tale of two “imaginary” friends. Let's call them Y and Z. Y watches info-tainment programs like “the Daily Show” and “Late night with Stephen Colbert” and those kinds of typical left-wing programs. Y does not watch Fox News or OAN, and if anything, he'll occasionally watch CNN. And yet, when I chat with Y, his political views border in Nazism. True story!

Z on the other hand mostly reads Conservative media. Z watches Fox news and reads conservative newspapers. Z also has conservative ideas on abortion, gay marriage, taxes and other issues. But, Z always votes for the liberals or the labor party, for pragmatic reasons. Z wants to maintain his medical benefits and retirement pension, and believes that the liberals offer more generous packages than the conservatives.

Problem is, in this day and age, many voters don't have landlines, and don't pick up phone calls from unknown numbers, much less pollsters.

I'll get back on AI some other day. Bottom line is we have so much data, but we don't really know how to use it yet.


    
Print - Comment - Send to a Friend - More from this Author

Comments(0)
Get it off your chest
Name:
Comment:
 (comments policy)

© Copyright CHAMELEON PROJECT Tmi 2005-2008  -  Sitemap  -  Add to favourites  -  Link to Ovi
Privacy Policy  -  Contact  -  RSS Feeds  -  Search  -  Submissions  -  Subscribe  -  About Ovi