Buzz about Artificial Intelligence
There is a lot of buzz around Artificial Intelligence. There are multiple platforms available. There are versions that will let you ask all sorts of questions or give it a lot of prompts to write text or code according to what you want. But AI is only as good as the average of data it was trained on. That means that those chat AI systems are "average systems" not "expert systems." AI expands people's capabilities. It does not replace people when you need an expert. AI is best used in combination with a trained human. As a society, we desperately need to learn how to identify when AI has generated political ads or other creative works so that we don't believe them.
Now "average" is often "good enough" in a lot of business cases. That means that we will see many, many people use these systems instead of hiring average creative people for logos, business images, web ads, and simple coding jobs.
The chat AI was trained on the web. These systems pulled in huge amounts of information from the web. However, that information included both expert information as well as student papers, student coding projects, and hobbyist programs. For example, one AI system was trained on the projects on GitHub and others used the programming examples from StackOverflow. A vast majority of projects on GitHub are student projects, classroom assignments, and many others contain bugs. The examples on StackOverflow are often vastly simplified in order to demonstrate a programming concept. A common note on that platform is that the reader needs to add all the checking and logging that their system requires in order to make that code example into professional work. Some of these examples were posted 10 - 15 years ago and no longer match the current version of the language or have defects that modern coding standards fix.
Code generated by those AI systems will be simple, lacking the checking and logging needed, and may include bugs that are known to put systems at risk. Professionals who use these systems to generate a first draft of the code need to know how to add what is lacking and to know how to identify and fix the defects that AI puts in. Bing just offered code that would work in many cases, but would be a bug in certain cases.
Similarly, the answers, text, and images generated are an average of what is on the web. One study asked questions of different chat systems around the world and found that the answers reflected the popular opinions from the different countries, not an "objective truth." People will ask questions and believe that the computer is correct.
One place where "average" is of great value is in education. These systems can be used to work with students and can have infinite patience when a student is struggling with a concept.
Artificial Intelligence systems are part of our future. We will have the benefits and need to know the limits.