Artificial Intelligence Both Good and Bad AI stands for artificial intelligence. We used to think it was a misnomer, but lately AIs have been doing some strange things. The latest I know about is an AI which has developed a new system of mathematics. The AI was developed to generate what is known as conjectures. These are statement the scientists believe to be true, but can not be proven beyond a doubt. The AI has been named after one of the greatest mathematicians of all times, Ramanujan. The AI is coming up with thousands of these conjectures using what it has learned. Some scientists might laugh at those who get nervous at this and call them uneducated, but we could be seeing the very start of AI doing things we don’t know how to do ourselves. Facebook got nervous and had to shut down some of their AIs. We all know there are plenty of engineers and such there who shouldn’t get nervous when AIs act, but this was different. The Facebook AIs invented a new language and began to use it to talk to each other and once the people at Facebook saw this and realized they couldn’t understand what they were saying, they felt they had to let the axe fall. We know AIs are better at certain things than humans and are great at repetitive tasks which might drive workers crazy. I am talking about doing the same task in a factory over and over all day and night. It might be putting a wheel on an automobile, putting a product into a box or even something as complex as welding a seam on metal. One of the things AIs do which is happening in the area of game play is they will play games against humans. They have been doing this since early in modern computer history. The task seems simple with some games but as games get more complex, more complex AIs are needed. Researchers now say game theory might be helpful in discovering alien life. The University of Manchester thinks so. They are suggesting a cooperative effort linked to game theory which would be used in not only discovering alien life, but intelligent alien life. There are areas where AIs would improve what we are doing now. One of them would be news reporting. AIs have access to the entire internet and could be programmed to be non-partial in the stories they pick out to report. That would be quite a change from what is happening today. Unfortunately, this would probably never be allowed because the news would not be manipulated to serve the purposes of one party or another. One of the greatest contributions AIs could make would be in the field of medical diagnosis. This seems to be one of the jobs AIs and computers would be perfect for. We could feed all the data into our software and the AI could sort it in a way which would allow it to bring up any possible causes of disease in a person, narrowing the possibilities and suggesting the treatments for each one. Some people would not like this and say it might be over-kill, because so many possible causes might be listed, but the list could be by percentages of chances with the most possible on top leading to the least possible on the bottom. There is a movement which wants AIs and the people who build them and use them to be legally liable for mistakes they make. Insurance companies argue there are now AIs placed in situations which might give them the power of life and death over humans which complicates insurance coverage payments in some situations. One instance I can think of is your car. Almost all the new cars have some type of AI in them. Take for example one system which is the dynamic cruse control. It enables you to set the speed and it will slow down or even stop if the car ahead slows down and stops. It is designed to start again within a certain number of seconds if the car in front restarts. The automobile company may not have built or even designed this feature so the insurance company might want to sue the maker of the AI and the builder along with the car company in an accident involving the cruise control which they believe the AI had a hand. The theory is already being tested in law, but not completely yet. There is an interesting theory about AIs which an article in Forbes has espoused and that is putting an AI in charge of something which could be dangerous in some way is the same as putting a human in charge who has no functioning left brain hemisphere because they claim there would be no real life decision making. I don’t know about that, because I do not see how you can compare a computer program with a human brain in the first place. The article seems to lean towards the AI not being able to handle an unfamiliar, unexpected or unusual situation. It seems to me that could also be a problem for a human. There are certain situations where there can only be a number of recognized outcomes. As for the others an AI might be better suited to be used in a warning and advisory capacity, such as running a nuclear generating plant. What is the biggest fear of allowing an AI to control things? It is the AI will realize it doesn’t need us and start a Terminator scenario. Truthfully that is beyond any programming being used today. No nation would want to build an AI that way, because they would also be wiped out. What they might want to do is build one to wipe out a specific country, which would be horrendous. I hope this never happens but it is sort of the scenario we had during the cold war. Not with AIs but with humans where both the United States and the Soviet Union had enough nuclear missiles aimed at each other to destroy those two countries. When we program AIs, we have to be very careful to let them know what the wrong decisions are and prevent them from reaching those decisions. It is important for them to be taught right from wrong, but unfortunately this depends on who is programming them. Humans don’t always agree on what is right or wrong so this may be carried on to our AIs. I think one of the biggest problems will be when AIs start to build AIs themselves.
|
