The Next Big Machine Learning Startup Will Look a Lot like VisiCalc
My inaugural post for Singular Distillation! I hope you all like it, I will be trying to keep to a twice-weekly schedule, if not thrice-weekly. Connect with my on Twitter < @vasaulys > or here on Medium.
Does anybody remember VisiCalc? They seem to be a causality of the early personal computing era. Yet they are arguably one of the more important software companies of history.
Way back in the dark ages of 1977, Apple put out the Apple II. Commonly referred to as the first personal computer, it set the standard for home computing by building a (relatively) affordable machine. That part most people know. What most people forget is that the Apple II didn’t sell very well at first. Piecing the numbers together from a source, it seems apple sold at most 10,000 units its first year. 
But in 1979, that all changed. VisiCalc was released. Although its “only a spreadsheet”, it was a real breakthrough because it gave people a strong incentive to buy a computer. Spreadsheets prior to computers were a pain to calculate, even with calculators. The magic with VisiCalc was that cells could be written as formulas that relied on other cells; if those relied-on cells changed, then the formula’s answer changed.
It was a simple idea, but it a powerful one. By 1982, Apple would sell 750,000 Apple II units and VisiCalc would port its software to other platforms, achieving a million in sales by 1986.
The power behind it, and why it was so crucial to the Apple II, was because it did something that was not previously possible. Well, technically it was possible, it you were IBM with a small army of MBAs. But VisiCalc, coupled with the Apple II, put this power in the hands of small business owners.
This is what Machine Learning needs. There are many many platforms and tools but no products or solutions. Companies are racing to build the next platform, but nobody’s racing to build the product. Somebody needs to be building the next VisiCalc. They need to be build the product that wasn’t previously possible. Machine Learning is unlocking potential not previously thought of as realistically possible.
So forget your new neural network standards and algorithms, show me your cutting edge can-use-in-my-life product. That’s what’s needed right now.