Source: Deep Learning on Medium
AI, Deep Learning, Machine Learning & NLP: An Introduction
A brief summary of four of today’s most important technologies
Artificial Intelligence, Deep Learning, Machine Learning, and NLP are all search terms that are popular as human technology continues to advance.
But what are these advanced technologies and how do they differ?
Artificial intelligence is a technology or computer system designed to function in a way that simulates how the human brain thinks.
The term is an overarching field of computer science that encompasses a broad range of categories, including natural language processing, machine learning, deep learning, neural nets, content abstraction, decision-making and more.
It was first coined by scientist John McCarthy at Dartmouth University in 1956.
It’s important to remember that AI is really composed of algorithms that are fed by data, and the results or recommendations it produces are entirely dependent on the data inputs.
The more biased the data, the more biased the outputs. And vice versa. Take, for example, Amazon’s failed experiment using AI for recruitment, which produced results that were biased against women due to the number of male CVs in its training data.
Or, MIT’s Norman experiment which created a cynical AI who ‘thought’ only about death after feeding it data from a particular subreddit feed.
Machine learning is a subset of AI which involves ‘training’ machines to ‘learn’ from sets of data, enabling them to draw insights and make predictive decisions. It automates tasks and finds patterns or anomalies, learns from them and creates new rules for next time.
Benedict Evans of Andreessen Horowitz writes in his blog that:
“machine learning lets us solve classes of problem that computers could not usefully address before, but each of those problems will require a different implementation, and different data, a different route to market, and often a different company.”
It enables us to automate a particular task, at a massive scale — something that would previously require many, many humans, if possible at all.
Deep learning is currently the most advanced subset of machine learning, and thereby a subset of AI, which intends to bring machines as close as possible to human levels of thinking.
According to MIT Technology Review, “The software learns, in a very real sense, to recognize patterns in digital representations of sounds, images, and other data” by creating an artificial neural network.
Image recognition and speech recognition technology fall under deep learning.
Natural Language Processing (NLP):
Natural Language Processing (NLP) is an element of deep learning that involves translating text or human ways of speaking so that a computer is able to categorize and make sense of it.
NLP is an example of data enrichment. Using this technology, an AI can extract elements from a piece of data — for example, a company name, a date, an event (for instance, an acquisition), links, and sentiment.
It also enables AI to analyze other forms of unstructured data — from video, to search.
Combined with semantic analysis, NLP AI can also look at the context to determine meaning from a sentence or data point.
NLP innovations are helping us to improve search in a way that uses language which more closely resembles the way a human thinks. As a result, it helps improve predictive search, suggests words or sentences in text and email, enables voice recognition, powers translation and more.
The AI Revolution is officially here. In order to remain competitive as we
forge on toward further technical advancements, it’s vital that every company embraces this new technology and implements change from the top down.
Those that are able to understand and speak about these terms will find themselves at an advantage as technology continues to evolve.