Not a week goes by that I don’t hear or read about the autonomous truck. A very hot topic of conversation in this industry for the past year or so. Clearly we have moved from the “if” to “when” question in the context of the driverless truck debate. With so much focus and attention on the autonomous driver and truck it’s no wonder that the emergence of Artificial Intelligence in other parts of the industry have gone virtually unnoticed by trucking executives at large. Now, in the defense of anyone not familiar with A.I. applications it is a nascent technology. The commercial application of A.I. is only very recently starting to be understood. In fact, the term Artificial.Intelligence is ambiguous in and of itself, not to mention a bit unsettling.
The more accurate term is Machine learning (M.L.). So what exactly is Machine learning?
According to Arthur Samuel, Machine learning gives “computers the ability to learn without being explicitly programmed.”
M.L. essentially does 4 things:
- Optimization. If you have ever used Google maps to navigate a trip you have used optimization for routing.
- Scanning retinas to determine your identity at the airport.
- Anomaly Prevention. Visa prevents fraud by denying card access when it suspects abnormal purchases.
- Amazon knows what to advertise to you because people like you (your segment) have purchased or shown interest in the same products.
Machine learning has been around for over 30 years but some major advances have been made recently. The most important change has been a by-product of M.L. called Deep Learning. Deep learning has only been made possible over the past few years due to some new major market forces:
- Cheap computing power
- More accessible data storage
- Major advances in algorithms
What deep learning does is it enables feature selection and model tuning. As an example, think about an excel spreadsheet. First, you need to decide on the column you would like to work on (feature selection) then you need to decide on the best formula or process to “attack” the data with (model tuning). Deep learning automates this. Google uses deep learning with WaveNet. WaveNet can fully converse with any human in any language, even if the it has never heard that language before. The key here is that WaveNet does not need to have heard the language before. It is a learning application so, like a human, it recognizes the patterns and applies its newly acquired knowledge to speak a new language. However, unlike a human, WaveNet learns a new language in nanoseconds! In a more practical use case, a Japanese insurance company has laid off 30% of its claims staff and is replacing them with an algorithm.
So why should you care? The advances in A.I., M.L. and Deep Learning are disrupting and doing a better job on many tasks currently being executed by humans. The challenge in the transportation industry is that a lot of the current technology may not ever offer M.L. based applications due to the nature and age of the products. The good news is that a new wave of companies and products are coming out of tech centres like Silicon Valley that are taking aim at the logistics industry. With trucking companies continuing to look for ways to push up on depressed margins; these new applications cannot show up quickly enough. I mean, what would your margins look like if 30% of the dispatching and/or order entry tasks were executed by algorithms?
Have your say
We won't publish or share your data