intro
In this article I will be going over the basic fundamentals of how AI is created, specifically the techniques used to create it. To begin lets understand.
How AI is made
There are many different types of AI generation that cause this question to range in how complex the functionality is. A basic definition would be continuously attempting to reach the desired functionality, with previous mistakes and algorithms guiding the end result to the desired functionality.
To achieve this functionality some of these techniques are used in tandem with each other.
Machine learning
Machine learning is what's used to learn and adapt to data without being specifically programmed to, an example of this would be face recognition.Deep learning
Deep learning uses Neural networks to learn patterns in complex data, an example would be speech recognition.Neural network
A neural network is a group of interconnected nodes that process information by referencing the external input. This information is relayed between these units until a output is determined based on the weight of the nodes.
To better understand, were going to go through an example to get a better understanding of how these techniques are employed together.
how image generation is made
There are two types of image generation when we think about AI.
There is generating a new image from a text input like asking for "a dinosaur with a jetpack", and generating a image off of a "prompt" like asking for "starry night in Banksy style".
This method is not true AI-generation since it uses pre existing images to create essentially a filter, Because of this I'm going to mainly focus the text input method.
This method uses two neural networks to achieve its functionality. One generates an image based off of the text input, this obviously uses machine learning to understand what the inputted text means. The second compare's the generated image to reference images, building an accuracy score till it generates an image with a high enough score to be sent to the user as a result.
This iterative process of creating and recreating till the end result is the desired functionality, is how most if not all AI computation is structured, in our next section ill discuss this more closely.
Automating tasks with AI
Task automation is possibly the biggest use case for AI in recent years. Huge advents in deep learning technology has made previously time consuming or dangerous tasks easily completed with AI. Since the variety of things we need, or want an AI to do are countless, there isn't a specific way to create and AI to automate tasks, but what is unanimous across these use cases is the iterative process of doing it over and over again.
An example we could think of is a AI that plays snake, by giving the AI the data of the controls and the basic objective. After making needed adjustments in-between iterations, like telling the AI to stay away from the boarder the AI should successfully reach the apple.
This process continues until the desired result from the AI is returned a high percentage of the time. As expected this is how variety of functionality from AI is achieved, through continues adjustment from the developer and improvements from the AI.
conclusion
I hope this article gave you a better understanding of what techniques are used in AI and the methods used to employ them.
To continue gaining knowledge on this subject, checkout the sources I used for this post. (:
Top comments (0)