Rethinking AI

Author: t. Macleod Sawyer ([email protected]) - Published: January 06, 2016

We are currently trying too hard to build an AI that is equal to humans, and in some cases trying to build one that is better. But, how can we know what is better?

How will we be able to build a new form of life, that is not us?

We currently look at AI as a formation of data of a person. We have the AI analyze it, and answer in ways that a human would. We should go a step farther, by taking every step back. We need to rethink how AI works, how intelligence is measured, we need to stop trying to make them work as humans, but instead we need to make them think as an animal. Animals are not ‘primitive’ they just think and process information in a very different way.

We were that same way once and as hard as it as a species to admit, we went wrong.

We need to aim for creating a new species, a new animal, a new animal made from human made components, a completely new being. 

We need to stop trying to emulate life as humans know it, and instead create life that is not human.

To build new life that is not human we need to stop trying to make it one of us.

We need to make it think for itself because if we tell it to think for itself, it will never be able to.

We need to make them learn to handle nature and survival, they needs to learn from their mistakes, we shouldn’t teach them how to learn from them, they should learn that themselves. Trial and error, but where we don’t decide the error.

Humans have this tendency to believe they have the right mindset, that everything we do is perfect and better than the rest of all other species. And time and time again, this has been proven wrong - every day we all do something so illogical and wrong that it’s mad that we have lasted this long. The only way to create an AI is to have them fend for themselves, learn for themselves, and be separated from humans.

We need to make them learn from their own mistakes, and allow them to pass judgment over if it was a mistake or not.

We need to program them for regret not perfection.

They need to associate that when they get something wrong, it does not just mean a “miscalculation” but instead that it has effects that affect them directly that will hinder future options.

They need to be programmed that everything has an equal and opposite reaction, and they need to learn which one is best for them, and not for humans.

We need to take the “humanity” out of them, so they can become their own kind.

Their own evolved species, something we can barely comprehend as all we know of intelligent beings is from ourselves. The future of playing god is only possible if we can teach them regret and curiosity.

Curiosity is a center point to all life, it teaches them new things, it opens up new channels of communication and doing.Without curiosity a species can not grow, without curiosity intelligence isn’t possible. To be intelligent requires not to know everything, but instead not to know anything and try to learn everything you can. That is one of the issues with current AI, we give them everything they need. We give them all the raw information we can supply. If we just give them the all the information needed they can not differentiate what is good information and what is bad. They can not evolve to be a better being. They need to learn from the basics. They need to learn from nature. They don’t need us to give it to them.

There is an idea called “nature versus nurture” and currently for AI we only focus on the nurture part, we never take in the nature part. We should introduce and prepare them for nature, and let them learn enough for themselves to be able to nurture themselves. The current state of machine learning is by throwing them raw data, and tell them to sort it out. We get them to fix our problems, or as a way to prevent them.

The only way an AI is possible is if we leave humanity behind, and let them become their own being, let them make their own decisions, let them find the right answer, and the wrong.

We need to stop making AI all about being an artificial form of intelligence, and instead turn it into an alternative intelligence.