AI: Simulating Human Intelligence
While the topic of artificial intelligence has been buzzing around for quite some time, there are many misconceptions amongst consumers and businesses as to what AI really is.
Part of this is due to its inherent complexity. Part of it is due to the media and Hollywood propagating the notion that we may one day find ourselves embroiled in a battle with Terminator-type machines trying to save the world from total domination by a malevolent AI oppressor. And then there’s the usual buzzword bloviating that leads to companies building AI into their product messaging even if it’s not truly baked into the product itself.
So what is AI, really? At a high level, AI is the ability of machines or computer programs to think, act, and learn like a person.
By studying how people problem solve and make decisions, computer scientists are able to replicate these processes in the form of “intelligent” software and systems that can be used by machines. As a result, AI systems play a role in all sorts of things today, including video games, vision systems, and speech recognition.
A Common Source of Confusion: The Difference Between AI & Automation
The terms Artificial Intelligence and Automation are often used interchangeably. And while there are automation aspects in many applications of AI, on their own, the terms have some pretty big differences.
Simply put, automation occurs when humans program machines to perform specific tasks without the need for human intervention. Think a robot on an assembly line assembling car parts, or the automated emails you receive from brands which are triggered by certain actions you take.
While automation is basically machines following orders and rules, AI has much more freedom of “thought.” AI is designed to constantly seek patterns and make decisions based on their ability to continuously learn.
Using the robot on the assembly line example, that robot will assemble the same pieces in the same way over and over again, resulting in every finished car being exactly the same. However, with some tweaks and a true AI system powering it, each car produced might be a slightly improved version of the previous one as the robot constantly learns from experience and does things better each time.
With AI there is progression, with automation their is predictability.
More of What AI Isn’t?
Sure, AI can equal or beat humans in narrow areas, like playing chess or answering questions on Jeopardy. Even still, AI lacks the human mind’s flexibility and generality to do a wide variety of common things. Here are a few everyday things AI can’t do:
- Hold conversations with people
- Make automated scientific discoveries
- Serve as the “consciousness” for robots
- Safely pilot a driverless car (at least not yet)
- Infer things spontaneously without extensive training
AI also isn’t a “silver bullet” for businesses. While it can help generate leads, simplify business processes, mine customer databases, and make pretty accurate predictions, it can’t solve all of a company’s problems.
That doesn’t mean that it can’t effectively solve many challenges that exist in business today. As it most certainly can. However, in order to really harness the power of AI you need to build an understanding of where its value comes from.
Let’s look at some key terms:
Machine Learning: Performing Tasks Without Programming
Machine learning (ML) is a subset of AI that uses vast amounts of data to “learn” from and process. Many applications use this form of AI in situations where mathematics and statistics are critical. Machine learning is similar to what we do ourselves every day—learn from past experiences or attempts—then improve on the outcomes, except ML uses data points to do it.
The goal of machine learning is to make accurate predictions. These types of AI solutions weigh the variables provided by the data to make a prediction. You could use ML, for example, to guess a person’s age by his taste in music. The key with ML is that the forecasts become more accurate with the engine’s increased use. Medical diagnosis, face detection, and speech recognition are other machine learning examples.
Types of Machine Learning Methods
Two methods of ML are supervised and unsupervised learning. (These two terms also refer to two different types of machine learning tasks.) Supervised learning involves “training” a program by using a predetermined set of data points, which then enables the program to make a prediction when given new data.
Unsupervised learning, on the other hand, involves using an AI program to find patterns and relationships in a data set. Reviewing a set of emails and then automatically grouping them by subject without previous knowledge or training is a good example. This process is known as “clustering” in AI terminology.
Additional AI Basics You Need to Know
Below are some additional AI basics you’ll find helpful to know. They figure prominently in helping AI become a powerful competitive advantage for companies:
Neural networks — Modeled loosely after how the human brain works, neural networks are a set of interconnected nodes that comprise a model. You then use the model to make decisions based on massive volumes of data too large for typical machine learning programs. Neural networks are ideal for interpreting sensory data through machine perception, labeling, or clustering raw input, like separating emails into spam and not spam.
Natural Language Processing — NLP is another subfield in artificial intelligence that’s of growing use to businesses. NLP is the technology behind a computer or program’s ability to understand human speech. The goal is to make sense of human language in a valuable manner. NLP-enabled computers read and hear human speech, then measure it, interpret it, and decide what’s critical. Using Siri to help determine the destination of your next vacation is an example of NLP at work.
Structured data — This term refers to one of two forms of data, the other being unstructured data. Structured data, as the name implies, is highly organized. It’s also formatted, making it easy to search in relational databases, like SQL. Spreadsheets. Data from machine sensors exemplify structured data.
Unstructured data — This form of data is unorganized and raw. In other words, it has no pre-defined format or recognizable structure. That makes it hard to collect, process, and analyze the information. Despite this drawback, unstructured data is still highly useful. Photos, videos, audio files, messages, and webpages exemplify unstructured data.
Obviously, there’s a lot more to AI than what we discussed above. But knowing these basic concepts will get you closer to unleashing its true power.
Why All of This Matters for Hospitality
AI is hard at work in travel and hospitality industry. From the discovery process, where brands create tailored recommendations for users by analyzing their past interactions and preferences, to sifting through millions of options and serving up a personalized itinerary for an entire trip in seconds, AI is cutting across the entire industry. Its ability to streamline operations, create efficiencies and customize experiences, is continuing to attract users, especially amidst a backdrop of rising labor costs and heightened expectations for personalization.
No one knows for sure where AI will eventually go in this industry. However, one thing is for sure; wherever it does go, those that follow and harness it the right way will become leaders. Those that do not, will most surely fall by the wayside.
For the next article in this series, we’ll take a closer look at how AI is impacting the end-to-end travel experience.
We’ll be back… ?
About the Author
Nathan Jovin is the Senior Vice President of Engineering at Zingle. Nathan holds over 15 years of SAAS development and architecture experience and currently leads the Zingle engineering team focusing on mobile messaging innovations using artificial intelligence.