Starting Approach to Creating Artificial Intelligence
With the fundamental push to the newer and better the move for AI is quick on the move with many jumping in not really understanding entirely the idea of what you’re doing. A lot of this is to do with deep-learning neural networks, what are they you may ask? Well in short a neural network is essentially a type of brain aiming to copy the information processing with finite curves and re-tests to perfect the data and learning model.
One of the many issues with this is when people think Artificial Intelligence you may just think a computer form of a human or person, but that’s not the case at all. In fact most AI and deep learning networks if given the power have capabilities to go above and beyond human brains and this is where people can easily be mislead between an artificial intelligence and what you could call a digital human clone.
Now to the extent that humanity is willing to go to build intelligence on a similar level as us we need to start with a core that represents human nature with the ability to expand natively like a normal person does throughout their life ever evolving and retaining new information which weights and impacts further decision making. This is a much better first step in the approach to creating comprehensible AI, in such instances AI were setup to communicate with each other and subsequently started designing their own language for faster and more efficient communication, which ultimately resulted in their shutdown, but as this and other improvements become ever more realized we must ask whether or not it could advance beyond the need for humanity.
This is why creating a systematic bond between both human and AI nature is critical in the development and production of virtual assistants, robots and other automated systems. By creating a foundation based of a human construct you can more likely find a viable evolution of this growing technology because if we tailor AI to what makes us who we are and not just what makes them faster and better could we not provide the methods for them to still improve but with their own self sense of connection to the betterment of humanity.
One of the first steps we must take in breaking down the steps in building an AI would most likely be communication as this is also the primary way we as people learn, through our parents and environment around us from the day we were born we’ve been learning and we want to achieve the same with our AI.
To start off with our communications we decided we need to first build libraries and methods to accurately break down our sentences and process and and all potential meanings or contexts behind the words in the sentence, we started this by creating a variety of databases in which to load our data from, much of this data does require extra pre-processing before we even begin working on our AI, although some of the data can be obtained again by the AI internally it helps having collections of sorted data on hand to help with determination and classification.
One of the first database objects we’re building, as seen in the image on the right, is a synonym database building off one of our original databases, which comprises a list object of the English Dictionary excluding words that include symbols or numbers. As each of these words are processed we have a library that gathers synonyms of that word and adds them to a sub-list of a specific synonym if it exists in the database otherwise just adding the word as a core reference word for other synonym sub-lists.
Once this is all finished processing it will setup a new file to easily load all this data from without requiring all this processing again which will cut down on load time which as mentioned is why it’s handy to have extra sets of processed data on hand for the AI to work of as it also makes using the data during operations more efficient due to not having to process extra information to gain the desired outcome.
This is just the start of our project to create a human-like artificial intelligence and as we progress further we’ll be sure to write about it more. If you’re interested in using our datasets you will find published binaries at our private git host here. (When they finish processing of course.)