Troy Lau, Group Leader in Machine Intelligence, Draper
The scene is familiar to any large organization faced with turning large mounds of data into actionable intelligence. The data could be stacked in boxes or locked in hundreds of computer files—and there‘s no obvious place to start. So a team is assigned to review the contents. Days go by, then weeks, and the hours pile up.
Machine learning was designed for just that sort of problem. Machine learning is a method of data analysis that automates analytical model building. It‘s a branch of artificial intelligence based on the idea that systems can learn from data, identify patterns and make decisions with minimal human intervention.
One in ten enterprises now use machine learning. Common applications include fraud analysis, market segmentation, computer-assisted diagnostics and sentiment analysis, according to a recent report by MCM Ventures.
Should you buy or build?
But the decision to buy or build your machine learning capability isn’t easy for some.
Let me try to sort out the issues.
Buying off-the-shelf tools offers a host of advantages. Vendors take over the tricky issues of integrating new machine learning and AI applications into an existing IT environment and training employees to use the tools. They also offer specialized algorithms for tasks like image recognition. But there are limits to the off-the-shelf approach.
For example, if a bank wants to determine the type of customer that will default on a loan, the training data to create the algorithms must be unique to that bank because its customer profiles are unique. The same applies in retail. How can a company understand the buying habits of its customers when it uses generic customer information?
While machine learning is fundamentally data-hungry and strong machine learning algorithms like lots of it, it can’t be just any data. The data needs to be specific to the organization and its data sources.
Off-the-shelf software can be the safer route, but its rewards can be limited, and it doesn’t offer the disruptive competitive advantage that machine learning has the potential to deliver.
Tips for adopting machine learning
Following are a few guidelines for bringing machine learning into your enterprise.
Go Open Source. There’s good infrastructure out there, and it’s getting better all the time. It’s far more sustainable to use open source tools than to build and maintain custom infrastructure.
Hire the Right Team. Your machine learning team should know how to build end-to-end machine learning solutions to real-world problems, by which I mean that they need to collect and curate the data, ingest it, explore it and clean it. They also need to write the model, train it, evaluate it, iterate on it and finally deploy it.
Use Good Data. Currently there are a number of ways that machine learning researchers and scientists can get access to data to train their machines. The best way is tap into a large number of free and publicly shared labeled data sets.
We’re in the middle of a revolution, where software can be trained, not merely coded. If your company spends all of its time on running the business, but neglects to use machine learning to streamline your processes and get good insights, then you may be missing out.