Why Artificial Intelligence is Like a Hungry Beast
Within the next few years, as we all know, every industry and every business will be fundamentally re-imagined and re-wired with artificial intelligence (AI). At the heart of AI is a machine-learning algorithm that teaches itself to recognize patterns and discover hidden insights in data. These learning algorithms, however, need to be fed with an uninterrupted supply of data for the magic to happen. It’s like taming a wondrous beast that will generate an enormous commercial value. The more and better data you have, the better it learns. That’s how Netflix’s algorithm knows your movie taste better than your spouse does; Amazon knows which book you will be interested in reading next; Uber knows traffic conditions better than any of the city governments.
As algorithms learn more about us, they are better able to predict our needs and act on our behalf. Every company needs an AI beast today. Feeding this beast/machine should become the central focus of each organization. That’s where most companies find themselves in an uncomfortable situation. Feeding the beast seems a mission impossible for them, as they are faced with two unsolved challenges:
- Data volumes have already gone beyond manageable levels. From our kitchen to the toilet, office space to the parking lot – they're all getting connected, with a promise to make our lives better and achieve magical things. A flood of data is being generated by sensors in mobile devices, cars, shoes, and virtually every other physical object. More data means it's easier to satisfy the machine's appetite. Right? But there is a catch. IT and business teams often find themselves between a rock and a hard place with these torrents of data. They are struggling to decide what to collect and store and the costs involved. It can be overwhelming. Using the new machines without abundant data is like starving your beast!
- “Dirty” data—inaccurate, incomplete or erroneous—makes the machine “dirty.” Lots of data doesn’t mean having relevant data. Collecting dirty data will do more harm than good to the business. Like oil, data needs to be “mined,” “refined,” and “distributed.” While it can grow fast in scale and value, it's no use if it lacks accuracy and quality. In fact, 90% of companies still aren’t keeping their data clean, which is costing them tons of money. Unhealthy food makes you ill; in the same way, dirty data will make the beast deliver incorrect or biased results, which are hurtful to the business. And all those who are hiring highly-paid data scientists will end up burning a hole in their pocket, as data scientists spend 80% of their time in collecting, cleaning and organizing data sets rather than leveraging the new machine.
On top of the above, business functions (sales, marketing, production and logistics, among others) often use their own systems and analytical tools and it’s a mammoth task for IT to integrate every one of them, and provide a single view of all data sources. With time, data siloes proliferate and run out of control.
Turning data into actionable insight with the help of the new machine won't occur by accident, but by establishing and managing a “data supply chain” across the business. For that, you must improve not only how much data you collect, but also its quality. To compete in the accelerated digital age, it's vital to reassemble your data collection and refinement strategy — in fact, it’s where the digital game is won or lost. In the next decade, the breakout companies will be those that become masters in consistently feeding this abundant data into the machine for actionable, and proprietary, insights.
Bottom line – don’t jump the AI gun unless you have addressed your data woes. The new machine will only get hungrier and demand more data as time rolls on, because that is the very nature of learning algorithms. It will be impossible for you to tame the beast.