Some people use the terms “artificial intelligence” and “machine learning” interchangeably, but they’re not the same thing. To put it simply, artificial intelligence is a broad computer science concept that encompasses the idea of machines displaying cognitive abilities. These abilities range from visual perception and speech recognition to decision-making. Think anything from Amazon’s Alexa to Hanson’s Robotics’ Sophia.
Machine learning, on the other hand, is only one of the applications (or subfields) of AI. It’s the ability of the machine, or technically software, to “learn” by extracting insights from data, without specific programming. One example is IBM’s Watson “collaborating” with a human to create a Lexus commercial.
The field of artificial intelligence
The idea of artificial intelligence dates back as far as ancient Greece and Egypt. In modern times five scientists, including MIT cognitive scientists Marvin Minsky and John McCarthy, are considered the founding fathers of the discipline, coining the term AI in the mid-1950s.
But even before them, several other scientists contributed to the idea. They included math genius Alan Turing, who suggested building intelligent machines to solve problems and make decisions using available information and reasoning.
In the past six decades, AI has had highs and lows, but today’s resurgence seems to have staying power. It’s largely because computing power has finally developed to the point of being able to handle massive amounts of data, making AI applications feasible.
The tech giants — Google, Amazon, Apple and the like — are the heaviest investors in AI applications, pouring billions of dollars into it. McKinsey Global Institute estimated that internal investment in 2016 totaled around $18 billion to $27 billion. AI technology is now used for everything from powering virtual assistants, shopping sites and chatbots to curating your social media feeds, analyzing medical lab results and enabling smart appliances.
Examples of AI fields
Besides machine learning, AI has several other branches or subfields. Here are two examples:
Computer (or machine) vision
Machine vision is the ability of the computer to imitate the human eye — that is, to “see” and interpret images. Computer vision technology uses image processing and pattern recognition to do simple things like read license plates and more complex things like helping surgeons in the operating room.
Natural language processing (NLP)
Natural language processing is the ability of the computer to understand and process human language, which is different from processing the structured data computers typically process. (In fact, human language is unstructured data). For example, the smart toy Hello Barbie uses NLP, together with machine learning and advanced analytics, to carry out a conversation with a child. Alexa, Siri, Cortana and Google Assistant are other examples of NLP applications. NLP is not the same as speech recognition.
Types and examples of machine learning
Based on algorithms, machine learning uses data as input to identify patterns and “learn,” then make predictions or decisions. Depending on whether the data is labeled (or classified) or not, machine learning can be supervised or unsupervised. Additionally, reinforcement learning allows the machine to learn by interacting with the environment, via trial and error, to determine the best outcome based on feedback.
Another term to know is deep learning — a machine learning subfield that is, essentially, about computers learning directly from inputs such as images or sounds, akin to how humans learn from experience. Deep learning can be either supervised or unsupervised. One of the earliest famous examples of machine learning was Deep Blue, IBM’s chess-playing computer that defeated Russian world chess champion Gary Kasparov in 1997.
(Interestingly enough, Kasparov beat Deep Blue the previous year, but IBM made modifications before making the second attempt.)
Some of the many ways machine learning or deep learning is used include:
- Fraud detection by financial institutions
- Self-driving cars
- Recommendation engines for services like Netflix and Amazon
- Customer service chatbots
- Facial recognition
Machine learning in cybersecurity
The high-tech, telecommunications, automotive and financial services sectors are leading the pack of early AI adopters. Of all AI investments, McKinsey estimates that machine learning has the lion’s share, with external investment around $5 billion to $7 billion (or about 60 percent) in 2016.
In a 2017 survey of 2,500 U.S. consumers, PWC found that 63 percent believed AI could “solve complex problems that plague modern societies” and 68 percent thought AI was important for cybersecurity and privacy. Indeed, cybersecurity companies have made quite a bit of headway in the last few years in applying machine learning. Now ML is an industry buzzword.
Since machine learning can process huge amounts of data from multiple sources, it can identify patterns — and thus anomalies — that the human brain can’t see. Machine learning also allows for automating certain processes, such as threat detection and response.
Some experts see machine learning as a disruptor for the industry. ABI Research forecasts that “machine learning in cybersecurity will boost big data, intelligence, and analytics spending to $96 billion by 2021.”
“We are in the midst of an artificial intelligence security revolution,” ABI Research Industry Analyst Dimitrios Pavlakis wrote in the 2017 forecast announcement. “This will drive machine learning solutions to soon emerge as the new norm beyond Security Information and Event Management, or SIEM, and ultimately displace a large portion of traditional AV, heuristics and signature-based systems within the next five years.”
Machine learning also has its drawbacks. One is potential bias, since machines are trained based on specific data characteristics — or at least in supervised learning, which is what cybersecurity solutions most commonly use today. In other words, the same context that the machine learns doesn’t necessarily apply in different situations.
Another drawback of machine learning is that the threat actors could use it as well to automate their attacks, although for now that doesn’t seem to be happening on a large scale.
What to expect next
While artificial intelligence is bound to greatly improve cybersecurity, don’t anticipate it to replace human analysts and engineers any time soon. But expect to see the industry continue leveraging AI to supplement the human element, detect threats and automate responses.
- Lexus Europe Creates World’s Most Intuitive Car Ad with IBM Watson, IBM
- History of Artificial Intelligence, Harvard University blog
- A Brief History of Artificial Intelligence, Live Science
- Super surgeons: Proprio aims to bring computer vision and AI to the operating room, GeekWire
- Machine Learning for Beginners, edureka!
- Types of Machine Learning Algorithms You Should Know, Toward Data Science
- A Short History of Machine Learning — Every Manager Should Read, Forbes
- What Is Deep Learning? 3 Things You Need to Know, MathWorks
- Bot.Me: A Revolutionary Partnership, PWC
- Machine Learning in Cybersecurity to Boost Big Data, Intelligence and Analytics Spending to $96 Billion by 2021, ABI Research
- Machine Learning: Disrupting the Cybersecurity Industry, Information Age