top of page

Computer Brains (AI/ML)

I’m always skeptical when a new tech fad comes along. Initially I thought Cloud was just a rebranding of outsourcing the tech stack (IaaS). I was wrong. The cloud tech disruptor is so much more. I still think “Digital Economy” or Digital Enterprise” is just a branding thing to sell consulting services. I don’t see anything new there that wasn’t enabled by natural, ongoing IT advancement. Blockchain to me is still way overrated as a disruptor outside of the finance world. I admit I could also be wrong own that one. AI/ML is something that I immediately saw as a major disruptor. I think AI/ML and The Internet are the biggest ones of my IT career.


I recall Andrew McAfee’s* presentation at MIT when he talked about Hippo’s vs Geek’s and how the Hippo’s will be crushed. Hippo = Highest Paid Person's Opinion. If you’re in IT you know what a geek is ;). He shares compelling evidence that number crunchers (geeks) using advanced analytics and AI/ML significantly outperform the Hippos in making optimal decisions. He even shows that the geeks in a different industry will beat the industry expert/Hippo. As a consultant having served many industries this is easy to believe as I’ve always believed Industry Knowledge is overrated.


We now have tools that fall into the AutoML category. Automated machine learning (AutoML) is the process of applying machine learning (ML) models to real-world problems using automation. More specifically, it automates the selection, composition and parameterization of machine learning models. Nirvana is when these tools start replacing the Data Scientist. Some claim they are already there but when we did a Proof of Technology (POT) on one of them this was clearly not the case.


That gets to two hurdles AI/ML faces right now:

  1. You need specialized labor to create functioning AI/ML solutions. These folks are very hard to come by. This is why so many AI/ML projects fail. Its the hippos forcing the use of the new tech without knowing what it takes to be successful

  2. You need good basic data practices. This is a big hurtle. There are still so many companies that do not have the needed data discipline and capabilities need to feed the AI/ML brain. It reminds me of the time when supply chain optimization tools came out (e.g. i2 in the earlier days). I was at a client where a hippo tried to drive a company to constraint based manufacturing planning and they couldn’t even get simple machine/work center capacity loading right. Garbage in - garbage out no matter how sophisticated the planning engine was.


So what this means for the average company is that you need to buy pre-built AI/ML solutions. Do not use base AI/ML services/building blocks and try to build an AI/ML solution by yourself. You will fail or you’ll spend way too much money for what you get. E.g. if you want to use AI/ML to analyze your service tickets. Have the machine determine root causes or predict service volumes. Then go with someone that has built that already. Another typical example are chatbots or more specifically AI Bots. If you want to use that technology to support your help desk go buy it from a firm that has already built this for 100s or 1,000s of help desks. Don’t do it yourself. Your bot will suck and piss callers off.


The future here is amazing and I can’t wait to see what kind of AI assisted solutions my kids will enjoy throughout their lifetime. I’ll be watching from somewhere warm, preferably, a beach.


* Andrew McAfee, an MIT scientist and bestselling author, studies how modern digital technologies are reshaping our world

10 views0 comments

Recent Posts

See All

Comments


bottom of page