DataDumb Things

Terrible Ideas and AI

Sometime around the beginning of the Industrial Revolution, the world hatched one of the absolute worst ideas mankind ever had:

How do I get a human being to behave like a machine?

It’s not one of those loud soapbox type ideas. Rather, it’s one of those quiet, pernicious, ideas that cut a deep trough through the workforce. So deep we are still struggling with it today. I imagine Henry Frick sitting with polished boots up on his desk twirling his mustache, thinking about how much money his machinery was making him. If only he could get people to behave the same. Wouldn’t Ford’s assembly lines move so much better if people could just stop getting sick? Stop procrastinating, gossiping and backstabbing? Just work more hours, damn it! Think of the profits if people could just stop being people and start being machines!

There is just not enough space here to chronicle all the exhaustion and pain that has come from this one awful idea. And yet here we are, working a 40-hour work week, on DST, carefully tracking our paid time off. Well, here comes an idea:

How do I get a machine to behave like a human?

Sam Altman at Y Combinator just laid it out there last week: “[With AI] entire classes of jobs will go away and not come back.” Ginni Rometty from IBM has been banging the Watson gong all month, talking about trillion-dollar markets and Watson on-premises. The list goes on.

Remember Watson on Jeopardy? That was 2011, friends….eight years ago! Shouldn’t we all be unemployed by our robot overlords by now? Listen, I’m going to be direct here: I’m not an AI contrarian. I absolutely love AI. It’s just that… you and I ….we’ve been to this rodeo before. Technology makes some really big promises, and those of us with the job of allocating limited resources have to deliver real results.

So if the pressure is on to do something with AI, here are three things to consider before embarking on an AI project at your company:

  1. Are you trying to make a machine behave like a human?
    If you think about it, this is the same awful Henry Frick idea. How do we get the human out of the mix? It’s is a dark pipe dream. Even if you are “successful” outcomes are very unpredictable If you are under pressure to do something with AI think about these things.
  2. Do you know exactly what you want AI to solve?
    Can you point to the exact problem you are trying to solve? And is the problem unlikely to be successfully solved by conventional means? If the answer is yes to either one of these? You have an academic or pet IT project on your hands. Expect poor outcomes and a huge cost.
  3. Do you have the goods?
    As Deming said: “In God we trust; all others must bring data.” When we talk about enterprise software a single system can be massively complex. SAP has over 70,000 tables of data! This is the one thing that prevents projects from getting off the ground. If you want to make your project a lot simpler, make sure you have a strategy for data procurement, transformation, persistence.

If you are struggling with getting large sets of complex data together drop us a note at We can help get a complex and diverse set of data together for analysis. Forget desktop toys like Excel. This is pure server based harmonization.