AI in the Real World
So far, I have only really addressed the concept of AI, but not truly explored the existing technology. As I stated in a previous post, if AI is going to replace a person at a task, it must be at least as “smart” at that task as the person who would perform it.
While I have no way to directly test whether AI can replace a worker at a specific task, I can test its function on other tasks. I found an online game of 20 questions that claims to be powered by a “neural network” form of AI.
What that means is that the computer learns to predict results after being shown results and different ways of getting to them. Another way to think of this is that it learns what things are as it plays the game. For example, it could learn “elephant” by playing enough times and would “learn” to ask questions like “Is it large? Is it gray?” However, it also learns from incorrect questions. If it asks whether it is wet, for example. I would say no and when the answer is reached, it would know not to ask that question, or how to properly use that information in the future.
Essentially this form of AI is just using statistics to predict results, but the statistical model changes with use.
Now onto the experiment side of things. You’re probably wondering how this could apply to the workplace. I’ll even admit that a game of 20 questions is a bit of a stretch in that sense, but it does give an idea of the current technology.
In the game I played, I was thinking of a cat. It predicted that correctly in only 18 questions. After the game, it even gave feedback on how it learned from the feedback of other users and how my input compared. Even with differences in information, the algorithm still guessed correctly.
While guessing I was thinking of a cat isn’t a replacement for a worker, it is indicative of ability. In situations where prediction is a component, humans may be replaceable. Rather than taking a job though, it may reduce the load on the existing workers. This ties into the idea from a previous post where I used an NCEP report on the progress of computerization.
What I can apply from that report is the idea that the progress is manageable. While a dramatic change from where it started, the application of the technology still is not enough to replace a human outright in many of the decisions that need to be made.
