Artificial Intelligence has been making waves in industry for years. However, there’s much that the average person still doesn’t know about it. In this article, we’ll review what artificial intelligence really is, and five things you should know about how it works in industry.
(7 minute read)
|Ask anyone what Artificial Intelligence is, and they’ll most likely quote Siri or another conversation software, like Alexa. They might also mention big data analytics, or advanced automatic computing. Essentially, these are both correct, but artificial intelligence is so much more than that.|
At its core, Artificial Intelligence is a computer program that learns. The major difference between this computer program and other programs that we design is that when we code, we know exactly how we want the computer to solve the problem. Through the design of an application or a function, we can dictate every step of the process. With Artificial Intelligence, AI for short, we aren’t sure how to solve the problem, but generally know what the solution looks like.
Could you describe exactly how you recognize a face? Let’s say a friend of yours has never met your brother. Could you dictate the steps required for your friend to correctly identify your brother, without ever showing them a picture of him? It’s safe to say it would be nearly impossible. However, you could probably teach someone if you kept showing them pictures of your brother’s face (and others), and told them when they correctly identified him. Eventually, your friend would learn enough about the traits of the face, and they would be able to identify it on their own. This is roughly how artificial intelligence works.
In the above example, the learning described was “supervised”. In supervised learning, we give the AI program many data points, and start showing it what the “right answer” looks like. Eventually, we ask the program to identify the “right answer” – be it a face, an outcome, or a spoken word. We “reward” the program for its correct answers, and eventually it starts to get more accurate solutions. In this method, we’re not telling the program how to solve the problem – we’re merely showing it all the data, and what a “solved” problem looks like. We “supervise” its learning until the program reaches a degree of accuracy that’s acceptable. What, then, is unsupervised learning? Giving an AI program a problem without a strategy or a solution. Unsupervised AI is a newer field of research, but promises to be even more powerful. For example, Google's DeepMind recently demonstrated an unsupervised AI program that was far superior than supervised AI programs - and any human - at playing the complex board game called Go. Another company, Cortica, recently demonstrated self-driving car that learned to drive through an unsupervised process.
Artificial Intelligence can help industrial businesses in two areas: process and insight. With regards to process, AI programs or AI-enabled robots can automate or replace the need for humans to perform menial tasks as part of a larger process. AI programs can also conduct analysis on existing process data and provide recommendations to tweak elements of the process to help it run more efficiently. Similar programs can analyze large datasets and provide insights on the business, based on any goal – be it reducing operational costs or transport time. This often happens by applying statistical analysis to different suspected relationships in the data – but much faster and more thoroughly than humans are capable of.
We’ve started to see three distinct themes of AI-based products based on the above capabilities: voice recognition, image recognition, and pattern recognition.
Voice recognition has historically been the most obvious AI-based product. From automated customer service on the phone, to chat bots on websites, to getting Alexa to tell you the weather, voice recognition software has been replacing our routine tasks for a long time. Image recognition has been largely contributing to a buzz-worthy product of late: self-driving cars. It has also been applied on an industrial scale by taking photos of warehouse setup and recalling where certain materials or products are kept. Image recognition can be used to identify people, locations, objects or substances almost instantaneously. Last, pattern recognition is the driving force behind predictive analytics AI products, resulting in better insight into the business or industry. For example, industrial shipping company Flexport uses AI to determine the most predictable delivery methods and routes.
Now that we’ve got a good idea of what Artificial Intelligence is and how it works, here’s 5 things any business leader in industry should know about it:
1 AI solutions continually improve. Because it “learns” from examples, as shown in our supervised learning method, this means AI solutions will only get more accurate. As time goes on, more data is collected and becomes available, therefore more patterns and “correct answers” can be identified. What else does this mean? Since most AI software is open source (Google, for example, makes theirs available to everyone), the company with the most data wins. More data on your industrial operations = more patterns for your AI program to identify = better insights and predictions on your processes and operations.
See how we get you more data to drive better testing insights for your company.
2 AI can be applied to many problems in many industries. Just as humans can learn almost anything, Artificial Intelligence programs can learn almost anything, and can be applied in any major industry. It’s already replaced the need to manually watch and trade stock, has been used in Walmart to keep track of inventory, and predicts drilling locations for oil and gas producers. In fact, McKinsey values the use of AI in the upstream oil and gas industry at $50bn. Imagine how it could affect the industry overall if applied to each step in the process, from drilling to trading to distribution. The possibilities are endless! Just keep in mind…
3 AI algorithms are highly specialized. Just because an AI system is good at one thing, doesn’t mean it’s good at another. In an example from this Harvard Business Review article, an AI program can quickly and accurately translate Chinese to English. However, it can’t identify what a Chinese character means – and it definitely can’t tell you the directions to the nearest restaurant in Beijing. These systems are first built with a single purpose in mind. It may take multiple systems or programs to achieve a “do-it-all” AI platform. This means that implementing AI into your businesses operations isn’t a one-shot deal. A system must be implemented for each problem there is to solve, or process to automate. This makes it important for business leaders to understand what parts of their operations are losing them the most money, and should be tackled first.
Use our online calculator to see how much money your business could be losing due to testing errors (opens in a new window).
4 AI derives insights from statistical patterns, not concrete rules. Artificial Intelligence processes have low “interpretability”, meaning it’s difficult to determine how they arrived at a solution. This can make a decision hard to justify, since you never saw the whole process. It requires a level of trust in software systems that most people aren’t comfortable with. Interestingly, this is almost a reverse of our initial example of describing facial recognition to a program. We could show it the answer, and tell it when it’s wrong, but we couldn’t describe the process. Similarly, AI programs can show us the data they used, and give the end result, but can’t describe how they arrived there. Business leaders implementing AI technology in their operations must be sure to accurately and carefully define the problem to their systems, in order to be confident they are getting the answer they need.
5 AI can increase productivity of workers, not just their bosses. The most prevalent fear in the adoption of Artificial Intelligence systems is that computers will replace humans, stripping them of their jobs and income. This can result in major pushback from field-level employees when an industry attempts to introduce this technology – causing the new systems to fail. AI and consequently, machine learning, don’t necessarily take jobs away from humans. In fact, AI can empower humans to do more of what we are best at: ask more questions and seek more answers, while eliminating task that are menial, repetitive, or distracting.
Business leaders risk being left behind if they don't start implementing Artificial Intelligence solutions in their industry soon. How can this incredible technology change your business?
Validere’s IoT hardware uses Artificial Intelligence to make BS&W testing and results painless, both in the field and the boardroom. Check it out here.