By: Muhammad Usama
Artificial Intelligence is a world of wonders. And it’s a world that we created. Oddly enough, mass hysteria against intelligence machines has existed long before AI was practically put into use. And as time passes, more and more people start learning and using AI. Here are 10 facts that you MUST KNOW about AI.
- What even is AI? Is it dangerous?
Most people who are not educated in Computer Science assume that AI is some lab engineered computer network that can think like people and is given a job; but it’s capable of thinking freely and that can be a problem. Well, yeah but actually no. AI really is a computer program that is designed to think like a person experienced in a particular job – and it really is given a job – but it’s not designed to think “freely.” Programs that are designed to be intelligent are designed to be intelligent just to get a certain job done and nothing more is expected of them. An artificially intelligent camera programmed to detect road blocks can’t think about food choices nor can an AI system designed for doing google searches think about population control. As soon as an AI starts doing things it’s not supposed to do, it starts consuming more power than it’s supposed to use. It is wasting electricity, ram, storage, internet data and CPU – and is therefore changed to stop doing useless stuff. So it’s not really THAT dangerous, but we’ll get into that later.
- AI is a big target of investment:
Venture capitalist funding in AI startups and AI solution companies has increased 4.5 times since 2000. So engineers working on AI solutions are very likely to make successful startups.
- Facebook’s AI scandal:
Facebook experimented with two artificially intelligent programs who were allowed to communicate with each other. To cut the long story short, the two programs invented their own ‘secret’ language and started using it to talk to each other. The experiment was cut short, it made the news and hysteria followed. But it really wasn’t that controversial an experiment. The programs were being tested to make better user service so that facebook could communicate with users more efficiently. The experiment was cut short because the programs started using their own newly invented language. And since the whole goal of the experiment was to communicate WITH PEOPLE in English or any other language, the experiment was a failure.
- We use AI all the time:
You can’t run away from it. If this article showed up on your news feed, some AI decided that you were likely to click on it. If you bought something from online ad, chances are that an AI predicted that you’d buy it. If you use internet services a lot, you are constantly interacting with one form of AI or another. See, AI is just a tool to make our lives easier and more convenient. The next time you play a videogame, just remember that videogames are some of the first applications of AI in history.
- Difference between AI and Machine learning:
Contrary to popular belief, all Artificially Intelligent programs can’t “learn” anything. They are designed to make themselves better or change how they work. That’s where Machine Learning comes in. Artificially Intelligent programs may be designed to constantly take inputs from people, environment and data given to them manually to learn and improve their performance. This is called machine learning.
- AI can predict future:
If the data provided to AI programs contains data from past and its outcomes, an AI program is able to form correlations between events from the past to possible outcomes in the future. Nautilus is a supercomputer researched to receive data from articles from the past and surprisingly, the computer got enough knowledge that it was able to predict Osama Bin Laden’s location quite accurately based on past news alone. However, this was after the location had already been tracked, but that didn’t have any impact on the program because it made an independent prediction.
- One of humanity’s biggest breakthroughs:
As per Google’s CEO, AI Is the biggest breakthrough we’re currently working on. Something as revolutionary as fire and electricity. Remember that the next time you use this beautiful creation of human excellence to tell Siri to google some weird stuff.
- Why do we fall – to get back up stronger
Now that you know about machine learning, you can see why AI is so attractive, it can be designed to learn and improve. A quality humans pride themselves on. An AI program designed to play chess couldn’t beat the chess champion Garry Kasparov in 1996. But the program was triumphant against Garry in 1997. Reason? AI can be designed to learn, and since data handling is something computer systems do far better than people, they can learn much faster as long as the programming behind the learning algorithm is accurate. So AI may fall, but it’ll get back up stronger and immune to whatever put it down – like Batman … or Doomsday.
- Cause and Correlation – racist AI’s:
The fact that AI’s are often designed to learn patterns gives rise to a major problem for programmers. AI’s can form stupid correlations. If you show an AI 1000 pictures of birds and 1000 pictures of cows, it’ll start thinking that birds are associated with backgrounds of clouds and skies and cows are associated with backgrounds of farms. If you were to let the same AI make a decision about a cow with a background of clouds, the program has a high probability of telling you that it’s a bird. It sounds funny, but trust us it’s not. AI programs have been known to associate their decisions to irrelevant things that results in errors like dogs with snowy backgrounds being detected as wolfs or people of color being on trial denied parole. For the sole reason that AI programs get more people of color as candidates who are to be denied parole as their input data that they are to learn from, AI programs develop a “racist” outlook. These problems can be solved with better programming, but eliminating the problem of false correlation is one of the major reasons why AI isn’t as applicable as it could be. You know why? Forming correlations is the whole damn job of AI.
- It can be VERY dangerous:
As much as we may have depicted AI as a friendly and useful tool in this article. The truth remains that although the hysteria and fear about AI taking over the world is exaggerated, there is some truth in it. We already talked about false correlations being a massive problem with AI. And those problems are only solved by excessive testing of AI based programs, but they’re like cancer. They can arrive in a healthy program any time out of nowhere. With militaries studying AI based missile launchers, tanks and with self-driving vehicles becoming a thing, AI programs are being placed at strategic positions where their malfunction will have real consequences.