The last few years of machine learning progress is shining a bright light for the future of society. We can now start off loading human labor to computers and sensors - the first step down a clear path towards abundance for everyone. New companies, attracting talent and resources with the promise of a better future, in an efficient market is how we chase this goal. It’s a shame that the vast majority of startups are failing society as a whole by pursuing technically timid and status quo business ideas.
Browsing through hundreds of ML & data science job listings here in LA is depressing. Almost all can be lumped into a general category of “squeeze a few points of engagement out of our users” - not valueless, but an insult to the rich ecosystem of societal progress that could exist. Take a look at the “Top 50” startups, and imagine they’re all wildly successful. Congratulations on the bottle of liquor you can have on demand - you’re gonna need it when you realize what a waste of talent and effort most of those companies were.
Go back to basics: food, shelter, security, and ask “where does minimally trained human labor play a role, and how do I remove that cost?” Abundance is only possible when labor costs approach zero, and the weekly fruit basket subscription service, even when brought to its logical conclusion, has little impact on society as a whole.
The reason these ambitious pursuits aren’t more common in the ‘land of disruption’ is because of subpar incentives. Tech loves praising a good fundraising round or exit, throwing accolades to the founders who find themselves holding a big check without considering the hundreds of millions of dollars, hundreds of thousands of man hours, and hundreds of unrealized better ideas that died in the process. This is a bad optimization function.
There is an infinite set of profitable business that can be machine learning driven. There is always that one extra percent of ad targeting you can improve. These, for the most part, are bad uses of your time. Value measured solely in dollars is not correlated to meaningful change (see everything done with ML at quant hedge funds).
Measuring utility is hard, and a “I’ll know it when I see it” approach is valid, but a discussion about how and what is socially (and financially) incentivized is important. Making things more complicated is that a lot of technology starts out as prohibitively expensive before it becomes widespread and cheap. Many useful things we have today were once derided as “toys for the rich.” Thinking about where a business model leads on an infinite time scale is vital for understanding how to affect large change.
Why does it matter? A useful cultural goal is one that gets society closer to peace and prosperity. The more people whose basic stresses are dealt with, the more human energy can be put towards goals like longevity. I’m a firm believer that we can live forever, but we need to stop making juice pressers in the meantime.
Uninspired startups slow the pace of progress, diverting already constrained resources to pointless goals. It’s frustrating, lacks imagination, and actively robs all of us of a better future. Build something that matters.