When people think of recent technological buzz words in the news, “Artificial Intelligence” has to be at the top of the list. A quick search on Google News looking for articles that mention that phrase pulls up over 90 million links!
For the general public, those words trigger visions from science fiction, where computers think like people and robots come alive. The truth is that the world is far away from this type of general artificial intelligence. In fact, some scientists feel that “General AI” is not even possible. Even if it is eventually feasible, researchers are only taking their first baby steps in that direction. In the best case, we are likely decades away from that dream.
If that statement is true, you may ask, then why is there all the hype? You may have read that Google paid $600 million to acquire an AI startup from the UK called Deep Mind that had not yet generated any revenue. Maybe you saw that a computer had beaten a Korean master of the board game GO using AI, causing China to develop a significant national AI strategy in response. Or you just started to notice how Netflix really seems to know what movies you will like before you watch them or how Google Translate is getting really good.
Something is certainly going on.
When most people see the concrete results of AI, they are really seeing the output of a concept called machine learning. Machine learning refers to a particular type of AI that learns by itself. Without getting into too much detail, machine learning commonly refers to computer programs that use a mix of advanced analytical models, adaptive algorithms, applied statics, and operations research to find hidden patterns in large data sets without being explicitly told what to look for or what result to output.
AI programmers typically feed known training data into a machine learning algorithm. They then tweak rewards and penalties in the algorithm until it outputs a result that they are looking for. A simplified example is, if you feed the program with many labeled pictures of animals and reward it when the output is the word “cat” but penalize it when the output is “dog,” you get an algorithm that can identify cats in a picture. The more data you give the algorithm, the better it gets. Once the algorithm is trained, you can show it new, unlabeled photos, and you have yourself a program that will predict if there is a cat in the picture.
Notice the word “predict” in the last sentence. The algorithm is not always correct. In 2010, algorithms of this type were wrong over 25% of the time for this task. Therefore, their predictions of if there is a cat in the picture or not was only 75% accurate. Since 2010, these programs’ accuracy has improved greatly through significant advances in algorithm development and large increases in inexpensive computing power availability. Now they are correct over 98% of the time – better than a human who will get it wrong more often.
Machine learning programs are essentially prediction machines. You feed them large amounts of data, and they make a prediction. Over time, these predictions are getting more accurate as the amount of data we collect increases, computer power gets cheaper, and algorithms get smarter.
Supply Chain Predictions
In the world of supply chain, we make predictions all the time. Demand forecasting is a clear example where most companies make predictions related to future sales. Amazon developed their own AI systems to support their demand planning teams. They claim that their machine learning forecasting models can make predictions that are up to 50% more accurate than looking at historical time series data alone. They are now offering their forecasting ability as a service to other companies.
Here are some other example predictions where various companies have implemented AI to improve:
Predicting Your Future
It is clear that Artificial Intelligence is starting to make its way into the world of logistics. Machine learning may already be impacting your supply chain now when used by your customers, suppliers, and vendors without you even knowing it.
I don’t think we need advanced AI to predict that you will be using machine learning within your supply chain in the near future, if not already. A recent survey of 500 supply chain managers found that 90% believed AI will transform supply chains for the better by 2025.
I think it is safe to predict that the other 10% will soon change their mind.
-Tom Bonkenburg, St. Onge Company