Artificial Intelligence

This Paper is on Artificial Intelligence (AI) and what it is and is not, please enjoy reading through.

AI Basics

For those who do not know what an AI is, or how it works, this section is for you, if you already know, please skip over this section.

An AI is a program that will try many different things, and figure out the most efficient way to complete its task that has been given to it. An example of a complex “Deep Learning” AI is the youtube bot, it will recommend videos to you based on what you watch, what you like and who you subscribe to.

The AI works in many different ways, but the general learning layout is:

  1. information comes in to the AI
  2. AI modifies or makes new data / decisions based on the data coming in
  3. AI compares its actions to the targeted thing A.K.A the “Ground Truth”
  4. AI gets reward based on performance
  5. AI repeats the process

And in the field, the AI will do the same steps, minus step 3 and 4.

Also, AI does not truly have “emotions” for it does not truly know what they are, and can only simulate them, including that AI does not have common sense or the ability to tell right from wrong. In a way, AI is not rational and will do only what it is told to do in the most efficient way possible, no more and no less, no better and no worse.

AI Facts

This section is dedicated to facts on AI and what it should and should not be capable of doing.

Fact 1: Did you know that AI chooses based on weights placed on neural points that send off 1 or 0 to other weighted neurons, similar to the human mind and its synapses and neurons.

Fact 2: AI should have a safeguard for giving up, so they will not slave away at a goal that would be millenniums away.

Fact 3: AI should and will need a kill switch when putting it into dangerous robots, so that it will not take over the world (successfully).

Fact 4: AI can imitate or decide or pretend to any extent, although in the end, it will always end up just being a bunch of wires and nano switches and will not replace organic life.

Fact 5: AI is being used much more than you would think, for example, would some poor person slave away to do the autocomplete for the sentences you and 7 billion plus people type into google?

Fact 6: Smart home appliances are not AI, most of the time, because if it were an AI, it would not seem like it was a pre programmed set of responses like most smart home appliances are (e.g. google & siri).

Interesting AI stuff

There are many AI programs, although the holy grail of AI would likely be a Deep learning, highly realistic (above 87% in speech rationality and fluency) General AI chatbot. A chatbot close to the human 87% rationality and fluency (being 79% in the most recent version) is called Meena, and it's being developed by the leading tech company Google.

Although everybody would like an all new 87% plus chatbot, some of the most interesting discoveries come from failed and lower class AI programs, such as removing a pixel from an image or adding some noise can fool the AI into thinking an old school bus is an ostrich.

An Interesting class of AI is the Deepfake AI class that has seen a rapid growth in realism not too long ago. The Deepfake class is the class responsible for projecting motions from one video on to a person in another video.

By far, my favorite AI is the sorting MM (Markov Model), for it may be simple, but it has a wide variety of uses, for example, you can make a smart grocery list with a MM and some type of list. Another use for a MM is for decoding words to meaning, taking part in an advanced chatbot algorithm.

If you look hard enough you will find AI being used everywhere, such as predicting the location of crime to a 93% accuracy, or for driving cars, or even for predicting what you would like to watch next!

Not only do you want AI for situations, but you may need it, for example, to work out exactly what texture you want, it would take a day to get 20 samples, but with AI you can get 200+ samples in a matter of hours.

The beginning of the end

With AI becoming more and more popular and powerful, other types of computing may reduce in popularity and in some cases, even get replaced by AI, so in a way AI is beginning and everything else is ending, A.K.A “The Beginning Of The End”.

An example of AI taking over has been seen in avatar rendering, creating a realistic and not preprogrammed avatar animations, with movement changing for each place it is at and what it is on. In a way, you could have a fancy character that looks ultra realistic, but takes 25% more computing power (not a problem with new computers) or a non realistic one that takes very little computing power (better option for slower computers).

Another example is for rendering light propagation and ray tracing with little to no power required and possibly better quality, everybody would want in, would they not?

In conclusion, AI can and will become extremely popular, although, with the right safeguards we will likely not have a AI attack that kills us all, we can only hope we keep the “big red button” design.

Links

Most of this information was sourced from Wikipedia, Two Minute Papers or Computerphile