Not too long ago, you had to know a programming language to be able to interact with a computer. Today, you can talk to digital devices using human language. Just say, “Alexa, what’s the weather today?” or “Siri, where’s the nearest coffee shop?” and you’ll get answers — usually accurate ones.
This transition has been made possible by artificial intelligence and its components: machine learning and deep learning. They’ve enabled computers, smartphones, TVs, and other electronic devices to comprehend human speech and acquire a wealth of other skills.
Other examples of AI at work are:
- Streaming video services such as Netflix recommending movies based on what you and viewers with similar tastes have watched
- Credit card companies alerting consumers when an unusual purchase has been detected
- Manufacturing companies using AI-driven predictive modeling to determine when parts should be repaired or replaced
The impact of AI is just beginning. Companies are aggressively investing in AI capabilities to increase efficiency, lower costs, and boost profits. In a 2021 PwC survey, 86% of respondents called AI a mainstream technology at their companies, although about 48% said a lack of workers with relevant skills was a concern in adopting AI.
Students interested in pursuing a career in this exciting field and learning the differences between AI vs. machine learning vs. deep learning should consider a computer science certificate program.
Although some might quibble over exact definitions, AI is the group of technologies that enables machines to learn and act like humans. There are generally two types of AI — narrow and general.
Narrow AI vs. General AI
Narrow AI is what is in use today. It is an AI program with a single purpose — movie and song recommendations, for example — achieved through a fairly rigid set of guidelines laid out in the programming.
General AI, on the other hand, is machines thinking for themselves, as depicted in science fiction movies and books. It is theoretical at this point, and not something we see in daily life. A general AI system would comprehend the world as a human does and be capable of carrying out multiple tasks. Your household’s general AI robot could cut your hair, feed your pets, mow the yard, and maybe even manage your finances.
Within narrow AI are technologies such as machine learning and deep learning, natural language generation, robotic process automation, and computer vision.
While narrow AI doesn’t boast the capabilities that general AI might have someday, it’s still getting a lot done.
In industry, manufacturers use AI’s pattern recognition capabilities for a variety of purposes, such as identifying when jet engines are due for maintenance or bridges should be checked for weaknesses. Drones flying along a pipeline can feed images to an AI system to determine if there are sections under stress.
Companies deploy chatbots — those pop-up windows you see on websites — to try to answer customers’ questions or settle issues.
In medicine, physicians draw on AI systems to scan images such as X-rays and mammograms to look for areas of concern. Such systems have proved to be better than expert humans in finding cancer cells, for example.
Machine Learning vs. Deep Learning: A Deeper Dive
When trying to understand AI, machine learning and deep learning are key concepts whose distinctions must be understood. They are similar, but deep learning can process much more data, and the data can come in various forms — from spreadsheets where information is labeled to more unstructured data like social media posts. In essence, deep learning is a more sophisticated version of machine learning.
Machine Learning Basics
Think of employing machine learning like riding a bike with training wheels. Progress can be halting and the places where you can ride are limited. But when the training wheels come off, your options expand and you can go just about anywhere.
In machine learning, a programmer feeds data into a computer and employs algorithms and statistical models to help it get better at processing the data. As the system repeats its tasks, it learns what works and discards what doesn’t. The system learns its way to an optimal answer.
A streaming music service, for example, makes note of what songs you request, what bands you listen to most often, and whether you prefer classical music or classic rock. It compares your preferences to those of other listeners with similar tastes to recommend something you haven’t played on that service.
Deep Learning Basics
Deep learning is a more powerful subset of machine learning that processes data through artificial neural networks, designed to emulate how the neural system of the human brain works.
In a deep learning neural network, data moves between nodes on the network. As the data travels from layer to layer, it is assigned a value. The data with more value keeps moving through the layers. The last layer is the output of the data with the highest value, which should be the most accurate answer, whether it’s what a cat looks like or what a stoplight means.
The more data that’s fed through the system, the better the answer, as more inputs are learned and considered for the conclusion.
An example of deep learning is a self-driving, or autonomous, vehicle. It collects information about other vehicles, lane markers, traffic signals, and road conditions through a network of cameras and sensors arrayed throughout the vehicle. It processes that information and uses the best information to steer, stop, and start the vehicle. Statistically, a well-programmed vehicle will make a trip without incident.
How Machine Learning and Deep Learning Power AI
As AI has evolved, the field has advanced to increasingly complex iterations. A previous AI method was expert systems, which worked through a series of if-then scenarios. Programming expert systems was time consuming and the results were not spectacular.
The greater processing power of computer chips and the increasing amounts of data available have powered the growth of AI through machine learning and deep learning. The major successes of AI in recent decades have been achieved primarily through machine learning.
Moving forward, however, deep learning looks to take over most AI applications. One major advantage it has over machine learning is that machine learning is more labor intensive because of a part of the process known as feature extraction.
In feature extraction, programmers make a representation of the raw data that machine learning algorithms employ to do their job. The process is complex, requiring more steps for adaptation and fine-tuning. Deep learning models, with multiple layers, essentially do their own feature extraction, learning the raw data representation by themselves.
Researchers are gearing their work toward deep learning. The number of research papers about deep learning placed on arXiv, a platform where scientific papers are posted, is about six times higher than it was five years ago, according to the Human-Centered Artificial Intelligence Institute.
Just as machine learning evolved into deep learning, deep learning will likely take on more capabilities and morph into another level of AI. Will it lead to a form of general AI? Some in the field do not doubt that general AI will be achieved someday; others have reservations about whether it can be accomplished technologically (or whether it should be, due to concerns that general AI could eliminate too many jobs or even threaten human existence).
Be Part of the Future of Technology
We’ve experienced significant changes in the way we live and work due to artificial intelligence, machine learning, and deep learning. Automation allowed by AI has saved businesses billions of dollars. As much as life has changed so far because of AI, it’s likely to change even more as the field develops.
The continued growth of AI offers opportunities for those with the experience, knowledge, and skills to lead the next generation of AI applications.
Students interested in boosting their qualifications for a career in AI should learn more about Maryville University’s online certificates in computer science, including our AI and data science programs and post-bachelor’s-level programs in big data and machine learning.