What started as a bit of throwaway snark on Twitter became an interesting conversation about startups and failure.
“Fail fast” isn’t about getting lots of practice at failing so you can fail more.
— Justin Warren (@jpwarren) May 29, 2017
It’s no secret that I’m not a subscriber to the relentless hype that typifies much startup coverage. I work with, and interview, lots of startups but I live outside the Valley and enjoy poking fun at the overly-earnest buzzword extravaganza that all too often substitutes for useful products.
That said, I’m not against the idea of startups per se. Trying new things is how progress is made. It’s trying old things in the same way that didn’t previously work and expecting to be lauded as heroes that annoys me. This attitude—common to startups and established enterprise companies alike—that ignorance is something to be lauded doesn’t sit well with me.
The idea of “fail fast” is supposed to be about quickly identifying when something isn’t working, learning from it, and then rapidly moving on to something better. It’s not about rapidly cycling through all of the well established failure modes out of ignorance and hubris. Congratulations! You’ve reinvented the wheel only with more corners and… no, wait, that’s a triangle. Here, have some stock options!
Learning individually by personal trial and error seems massively inefficient to me, but maybe I’m wrong.
— Stilgherrian (@stilgherrian) May 29, 2017
Oddly enough, I had a conversation about learning with the CEO of an AI related startup Bonsai, Mark Hammond, just this week. Instead of throwing huge amounts of data at a machine-learning algorithm and then patiently waiting for it to do the equivalent of rediscovering calculus from first principles, the Bonsai approach is to teach the computer how to solve problems with its array of modern machine-learning tools.
That we can shortcut individually learning everything from scratch should be a fundamentally accepted idea in the modern world. It’s why schools exist. It’s why we value the expertise of someone who has spent years learning a particular trade, profession, or craft over someone who only just discovered that integers exist. Or at least why we should.
@jpwarren Exactly. It’s about recognizing failure early and having the foresight to let go of those sunk costs and move to the next thing.
— Bill Plein
This article first appeared in Forbes.com here.Tags: ai, bonsai, fail, failure, ibm, machine learning, ml, startup