The power of artificial intelligence (AI) and machine learning (ML) is growing every week. These technologies are the future of procurement, and we want to help you understand them so you can pave the way for automation.
- the technology that allows machines to simulate human behavior and make decisions.
- the computer learns from data, often through pattern recognition.
Remember the old saying, “Red sky at night, sailor’s delight?” People hundreds of years ago didn’t necessarily understand the mechanism of high-pressure systems filling the air with dust particles that made the sunset appear red. They simply noticed the pattern—when we have a red sunset, the next day is usually pleasant weather.
And that’s what machine learning does. It figures out the patterns in data much faster than we humans could. When we get a computer to recognize a pattern and make a good choice, we can call that AI.
How Machine Learning is Like Board Games and Parenting
One way to understand how these technologies work is by using the analogy of the different ways we have of instructing computers.
Traditional software is like the game Snakes and Ladders.
- It’s a game of chance with no strategy involved.
- We can tell the computer exactly what to do, and it can play the game correctly.
Traditional programming is also like having a baby.
- You have to do everything for it. If you want the baby to go upstairs, you pick it up, and you take it upstairs.
- We can and must help the baby, and the computer, do its thing.
In the middle of the spectrum of computing is a game like Checkers.
- Checkers has simple limited choices to make.
- We set rules, and the computer decides how to apply those rules.
The middle of the computing continuum is like having an elementary-aged child.
- When children get older, you lose a bit of control.
- You have to communicate by setting rules.
- The computer gets to make some decisions based on the rules.
Advanced machine learning is like Chess.
- Chess has complex patterns and nearly unlimited choices for the player or computer to make.
- We have to release more control to the computer and give it more flexibility in the things it tries and the ways it looks for patterns.
Advanced machine learning is like having an adult child.
- As children grow up, it’s no longer a great strategy to give them rules to evaluate every situation.
- We have to release control and allow them to make decisions based on their own experiences.
- The computer gets to make decisions based on the data it learns.
The trade-off with machine learning is between control and complexity.
The more rules we ask the computer to learn, and the more decisions we let it make, the less control we have, but the computer’s ability to handle complicated systems goes up.
At Tealbook, we’re using some models with more than 100 million parameters. You could think of that as a computer having a rule book with 100 million rules telling it what to do. That’s too many rules for a human to keep track of, so we let the machine learn these rules by giving examples of the types of situations we want it to handle. That’s what machine learning is. It’s the computer learning what to do from examples.
We use these tools, from traditional software to complex machine learning, and a lot of data in the work that we do building our supplier profiles. It’s not a question of one being better than the other. It’s a question of choosing the appropriate approach for the problem at hand.
For LinkedIn: At Tealbook, we’re using models with more than 100 million parameters. Think of that as a computer having a rule book with 100 million rules telling it what to do. That’s too many rules for a human to keep track of, so we let the machine learn these rules.