Following the herd can be costly.
“Common knowledge” is meant to be time-saver: all who came before you have already sorted out the best way to go, right? So by doing what everyone else is doing, by following the the paved road, you’ll spare yourself time.
At least, that’s the story.
The problem is that common knowledge doesn’t always live up to its name… so the real time-saver is knowing when to ignore it. Companies lose plenty of time, effort, reputation, and money on their ML/AI projects because they overvalue what everyone else is doing, while undervaluing practices that actually yield results.
You’ll certainly recognize this concept if you’ve read the Michael Lewis book Moneyball, which recounts the story of the 2002 Oakland A’s. The team uncovered undervalued traits in baseball players – that is, they found a way to break from common knowledge about baseball recruiting – and built a winning team on a small budget. By comparison, the teams that adhered to common knowledge poured money into players that scored high on metrics that, based on statistical analysis, had no correlation to real baseball wins.
If you want your company to run a winning ML/AI shop, then you need to think like the Oakland A’s and break from common knowledge.
Rather than make a list of common mistakes in ML/AI, I’ve made a list of undervalued practices. Following these practices can increase your chances of success, widen your opportunity space, and reduce risks.
I’ll share these undervalued practices over the new few posts. I’ve grouped them as follows:
- Part 1: Getting started with ML/AI
- Part 2: Hiring and Team Structure
- Part 3: Planning projects
- Part 4: Project Execution
If you feel you’re spending too much on ML/AI but aren’t getting enough back, read on.