Securing the right data and choosing the right model are two of the issues that companies could face when implementing AI

Many companies are pursuing projects to harness the of artificial intelligence (AI). While the is starting to make its mark in the supply chain domain, practitioners should not underestimate the amount of preparatory work that needs to be done to provide a firm foundation for AI applications that can yield real value.

AI was one of the topics discussed extensively at the MIT Center for Transportation Logistics 208 Crossroads conference, which took place at MIT on April 17, 2018. The discussions covered some of the issues that companies need to be aware of when developing AI-based supply chain . Here are some notable examples.

Do you have the right data? The basic transactional data that you want to mine for intelligent insights into a problem may or may not be adequate for the goals you have set. Widening the types of data used – something that is becoming easier thanks to the growth of Internet of Things technology – will enrich the data set but also will introduce more complexity. Also, data sets must be organized in a certain way for machine learning model building which takes time.

Which model should you use? There are numerous machine learning models and choosing the right one for the application you have in mind can be a challenge. Fortunately, this process is becoming more streamlined. For example, the MIT for Information and Decision Systems is developing ways to automate machine learning model building and has already achieved some notable successes. Moreover, the research aims to make it easier for mere mortals in operational functions to understand and participate in the process.

Does your team trust the results? When a new-fangled and much-hyped technology such as AI spits out results, there is no guarantee that practitioners will believe and act on the findings – especially if they are counter-intuitive. Some companies pursue relatively small, straightforward machine learning projects first to create trust and credibility. Pairing data scientists with supply chain professionals can help to break down perceptual barriers.

What about customers and suppliers? You might have to do some work to convince trading partners to join the party. On the other hand, some companies report that their suppliers initially drove the pursuit of AI solutions in the supply chain. Either way, AI projects bring opportunities for collaboration in key areas such as data collection.

Be aware of emerging challenges. The security of machine learning systems and processes is likely to become more important. For instance, criminals might attempt to mis-train systems for illicit purposes. The robustness of the technology is another issue, especially in safety-critical applications such as autonomous vehicle operation.

This list of issues is not exhaustive and new ones will undoubtedly emerge. AI advances coupled with the increasing accessibility of the technology – the basic building blocks are becoming available to a wider community of users – will drive more creative applications of the technology over the next five years.

This post is excerpted from an article titled Rushing to Adopt AI? Watch out for , by Ken Cottrill, Global Communications Consultant, MIT CTL, and published in the July/August 2018 issue of Supply Chain Management Review. 


Source link
thanks you RSS link


Please enter your comment!
Please enter your name here