Training My Artificial Replacement

I’ve been programming for 20+ years. I’ve made a career out of training my (technological) replacements to do my job for me, and lately I’ve read a lot of articles talking about AI/machines taking jobs from people.

Perhaps the biggest example of automation that I’ve seen is Amazon Web Services. IT Projects that 10 years ago would have taken 50 people 6 months and millions of dollars of infrastructure can today be completed in minutes by one person feeling frisky enough to spend next Tuesday’s lunch money.

Yes, jobs will continue to be automated - continuing the 500+ year trend dating back before the printing press. The arrival of AI and It’s synonyms make the current trend more interesting, but AI still has plenty of limitations. 

On the one hand, I’ve yet to see anything that gives me the impression that AI is going to start creating and running businesses any time soon. On the other hand, the future looks increasingly likely to hold a lot of smaller teams at smaller companies accomplishing bigger things.

Business: The Land of 10,000 Assumptions

Nathan Rupert “Confused or Disgusted?” September 11, 2015 via Flickr; 

Creative Commons 2.0 Generic

Telling people that I’m a data scientist is kinda fun. At the time of this writing, it’s still a relatively new job title, and it’s not one that most people understand. Once I reassure people enough get past the confused stares, I usually move into this “parable:"

Every business is a collection of approximately 10,000 assumptions. Said assumptions represent the Highest Paid Person’s Opinions (HPPOs) of said company’s leadership. Most of the assumptions are correct, or the company would go out of business. In most companies, It’s probably safe to assume that 5-10% of those 10,000 assumptions are wrong, it’s just that nobody in the business is sure how to tell the right assumptions apart from the wrong ones.

In the future, I firmly believe that the companies that most aggressively, (and efficiently) hunt down, identify, and fix the wrong assumptions will have a distinct competitive advantage over the companies that don’t. 

How do you discern which assumptions are right? You use the scientific method. You ask questions about your customers, you look at your company’s data to find the answers, and the answers almost always lead you to better questions. Lather, rinse, repeat. It’s data-driven science, that usually gets shortened to Data Science.

That’s what I do.

Stuff Siri Says

Siri is Apple’s lovable digital assistant that comes complete with just a touch of sass. (Or if you’re an Android/Microsoft person, Google Now and/or Cortana do roughly the same thing.) Siri represents a significant investment from Apple over a number of years, and is among the world’s leading efforts in Artificial Intelligence and it’s synonyms.

Despite her best efforts, Siri doesn’t always get it right. There are (sometimes NSFW) web sites that focus on funny mishaps that Siri makes. Among other things, this is a good case study in the limitations of cutting edge AI. 

Whenever I see someone talking about ultra-intelligent machines pulling an I Robot/Matrix/Terminator style takeover of the world I usually smile. Then I wonder how machines are going to pull that off when Siri can't even tell if she's screwing up trying to transcribe my text messages. Kinda like that time I told Siri "text my wife on my way home period"