How does AI work?

How does AI work?

Too many people are misusing the term ‘artificial intelligence`. AI is more than what you see on the apps on our phones, the smart tech in our homes or the personal assistants like Alexa or Siri that respond to our every command. AI machines don’t sit idly awaiting instruction, they “think” for themselves. No wonder people get it wrong; the very idea is enough to boggle the mind.

To understand what it’s all about, we need to set aside what our brains might interpret as AI and dig deeper into its core—the machine learning that is capable of truly amazing things.

Machine learning is the field of computer science that deals with algorithms that make predictions and learn from the data, without being explicitly programmed. Simply stated, the algorithm is let loose on the data. It learns from the data and makes predictions previously unseen. And the predictions become better over time as they are fed more data.

Here, two elements are essential: data and the right algorithm. The more data you have, the better. Limited and incomplete data increases error margins exponentially. If you give the machine garbage, it will merely perfect your garbage. As for algorithms, you must select the ones that are appropriate for your problem. As an analogy, to clean your house, you might use a vacuum or a broom, but you wouldn’t bust out a shovel and start digging.

Algorithms aren’t new, they can be traced back millennia to Greek mathematicians such as Euclid, the “father of geometry”, who developed an algorithm that became the most efficient method for computing the greatest common divisor (GCD) of two numbers. But now, with vast data and robust computing power, the machine can use algorithms to do what was once restricted to the human, but on scale and at speed.

Machine learning algorithms divide into two main categories: supervised learning and unsupervised learning. At the “supervised” extreme, the algorithm is presented with pre-classified data. Here, the goal of the algorithm is to learn the general rules that connect the inputs to the outputs. In supervised learning, the algorithms try to model relationships and dependencies between the target output and the input features.

For example, the algorithm could be presented with your employee data, including results achieved (in this case, output from the hours worked). Given that records of employees’ work are generally trackable via email and calendars, you can use supervised learning to determine what features are most impactful. You can then use these features to predict future performance and productivity. 

At the “unsupervised” extreme, no labels are given. Instead, the algorithm figures out the structure and pattern of the inputs on its own, by mining for rules, detecting patterns and grouping data to help derive meaningful insights. This is particularly useful when you don’t know what exactly you’re looking for!

All four types “learn” a target function that best maps input variables to an output variable. The relationships are then used to develop models, which in turn predict future actions designed to reach target outcomes.

Algorithms capable of unsupervised learning can identify the relationships between variables in a historical data set and infer its structure. They can also carry out optimization—a mathematical approach to problem-solving. As the name suggests, optimization searches for an optimal solution. It does this by starting at one point and systematically moving to the neighboring solutions until it finds the optimal option. These algorithms represent the world in a format that a computer system can use in order to perform complex tasks, and are instrumental in enabling signal processing, computer vision, speech to text, and natural language processing.

At this point, you may be thinking, “I’m not responsible for IT and don’t influence how my company uses AI.” That’s fine, but I can tell you that one of my biggest gains from working with machine learning for leadership impact has come from using algorithms.

Whether you love or hate math, learn like a machine—think through your work algorithmically.

要查看或添加评论,请登录

Chris Zechmeister的更多文章

社区洞察

其他会员也浏览了