An Intro to AI
Before we start learning about the different types and applications of AI, we need to learn what AI is. IBM describes Artificial Intelligence as “any human-like intelligence exhibited by a computer, robot, or other machine”. AI can be easily explained as three definitions: Machine Learning, Deep Learning, and General Artificial Intelligence.
What is Machine Learning?
Machine Learning is done through neural networks with at least 1 hidden layer. There are many different types of neural networks. Some of the most popular neural networks include CNNs, RNNs, MLPs, and GANs. While these are all types of machine learning, they all use very different methods. There is also some overlap with the applications that these networks can be used for. Most of the topics in AI are subcategories of each other. For example, Machine learning is a subtopic of AI, and Deep Learning is in turn a subtopic of ML. Keep in mind that I am mentioning different methods of machine learning, not naming specific algorithms.
CNNs
CNN stands for Convolutional Neural Network. CNNs are commonly used for image recognition, video analysis, and natural language processing (think Siri or Google Assistant). Amazon also uses CNNs in the “you might also like” section.
RNNs
RNN is an acronym for Recurrent Neural Network. RNNs are used for Face detection, prediction problems, generating image and video descriptions, text summarization, and even music composition! RNNs are also sometimes used for stock price predictions.
MLPs
MLP stands for Multi-Layer Perceptron. MLPs are known as the “vanilla” neural network, acting as the default for people who are learning to develop AI. MLPs were originally used for speech recognition, image recognition, and language translation. Currently, MLPs are mostly being by researchers. This is due to the fact that there are many more efficient and simple neural networks that can be used for the same applications.
GANs
GANs are normally perceived as the “scary” part of AI (other than general AI). This is largely part of their applications in deepfake technology. GAN stands for Generative Adversarial Network. A GAN is basically a combination of two neural networks. The first neural network is called the Generator. It creates output data. The second neural network in the GAN is called the Discriminator. The Discriminator takes two inputs from the Generator and must classify which is real and which has been fabricated. The job of the GAN is to fool itself into thinking that the generated data is real. GANs are currently used for creating music, art, deepfakes, or even make graphically intensive games! Nvidia’s DLSS is in fact a GAN, taking input of frames from video games, and trying to upscale a lower resolution version of the frame to look the same.
What is Deep Learning?
According to IBM, Deep Learning is “a subset of machine learning, which is essentially a neural network with three or more layers”. Deep learning is better at some tasks than “classical machine learning”, especially Unsupervised Learning. Unsupervised Learning is when the AI looks for patterns in unlabeled data, lessening the amount of pre-processing that needs to be done to the data.
What’s the deal with Artificial General Intelligence?
AGI is the proposed ability for an AI to understand and learn any task that a human can. AGI is widely talked about in science fiction. Many large figures in the tech industry have spoken out about their fears about the conception of AGI. These include Stephen Hawking, Elon Musk, Steve Wozniak, and many more.
Conclusion:
Although the listed Neural Nets are currently the most popular, there are always new approaches that can be and will be found. In his book “The Singularity is Near”, Ray Kurzweil talks about “the singularity”, a point in time when humans will be able to merge with AI, transforming our society permanently. This creates many ethical and philosophical questions that we simply cannot answer yet. AI is being adopted in nearly every field, and we need to make sure that we are careful about regulating it’s use.
