AI as Kinetic Energy: The Buildup

April 7, 2017

PUBLISHED BY Nathaniel Krasnoff

In my first article, I chatted about what Wildcat sees broadly in the AI space, how we think about AI and its applications, and how we acknowledge that an undeniable shift is already underway so we should figure out how to leverage it instead of working against it.

Before I dive into the nitty gritty for this post, I wanted to give a shoutout to the people who reached out with questions/comments on Part One. Keep them coming. I haven’t bitten anyone in at least a few months so no need to be afraid. For the folks who reached out with questions/comments from the last round, I posted some responses at the end.

______________________________________________________________________

Now that that is out of the way, this week I’d like to change gears a little bit and talk about some macro trends in the economy and what AI is doing to really change the topology of hundreds of different industries. In the context of Part One, if the hammer is already dropping think of this like the buildup of potential energy that led to hammer’s motion being triggered.

I was reading an article* recently that explained something to me about job growth and the macroeconomic underpinnings behind it that I had never learned before as a recovering engineer. The key point was that when the jobs report says we grew 300,000 jobs this quarter, and you’re reading that saying “well that’s not that many jobs for a country of 300 million people”, that’s the NET job growth in the country for that quarter. For e.g. If we created 5.3 million jobs and lost 5 million in the economy the net of it is 300,000. You’re probably reading that going “That’s so obvious, Nathaniel”, and that’s okay, but 1) I found that interesting so no need to be disparaging random people on the internet and 2) I didn’t know what the weight of that net was until I started getting serious about AI.

*Couldn’t find the original article, but here’s a Business Employment Dynamics Study from June 2015-June 2016 released this past January to give you a similar flavor.

This series is meant to serve as a contrast to the cup half empty view that it seems like many high profile folks are taking toward AI in the context of examples including, but not limited to, displacing the middle class. I am instead looking at AI for the opportunity it creates for people to move beyond jobs that requires no flexibility in thought to ones that leverage their creativity. Before we get there though, I do have to put on a scary face for you for a bit because the threat is not that AI will replace all the people, at least in the short term. “The threat is it will increase productivity rapidly enough to replace 20%+ of workers quickly enough that new jobs won’t be created fast enough to offset the losses.” That comes back to essentially the net I mentioned earlier literally falling out from under us.

I’d like to put some numbers behind this to show you what that actually looks like.

A report by the World Economic Forum (WEF) is predicting that over 5 million jobs in 15 major economies could be lost by 2020 due to “redundancy, automation or disintermediation,” and the greatest of these losses coming from “white-collar office and administrative roles.” This job loss is expected to be offset by 2.1 million jobs created, but those will be in focused “job families” (read math and science). Source

McKinsey found that one-third of the time spent in the workplace involves data collection and processing. Combined, both tasks have a 60% likelihood being automated out. The impact that these activities have is staggering as they make up 51% of the US economy accounting for $2.7 billion in wages. The report also highlights that by 2055, automation will affect some 1.1 billion workers worldwide correlating to $15.8 trillion dollars in wages. Source

Way back in 2013, Oxford University researchers predicted that 47 percent of U.S. jobs were at risk, and since then AI has improved much faster than we originally anticipated. Source, Source

Figure 1: 2013 Oxford Study: The Odds of Automation by Category

Between Google’s AI smoking the top 60 Go players in China to CMU winning out in Texas Hold ’em, which I can only speak for myself when I say that I wish that algorithm existed when I went to CMU to help subsidize the cost of my tuition, and 57 other really impressive applications of AI, the technology has moved much faster than people were expecting. Luckily, according to the same aforementioned McKinsey study, right now only 5% of these jobs could be automated outright, but given the nature of the breakthroughs above that percentage will rise rapidly in the coming years.

The push for this is coming from three sources. Academia motivated to prove that these problems are solvable, but the fire is really being stoked by VC’s like me seeking out AI companies building these breakthrough applications and big corporations paying high-six to low-seven figure contracts for the best PhD’s. The corporate side will be responsible for dropping the proverbial net out by displacing as many people as possible, which in turn, at least in the context of startups, would theoretically be best set up to create massive venture returns, which is what they pay me to find.

After months going through the mental gymnastics having looked at all sorts AI companies, two things came out of it for me:

The ethical dilemma of the technology

The observation that a lot of these technologies were being built by what my GP’s would call “Product-preneurs,” aka teams that are on the cutting edge, but aren’t going to turn technologies into big businesses. If you remember from Part One, these are the teams whose technology will cap out as “features,” which, for better or for worse, makes them uninvestable as their individual impact on the market will be mitigated by their inability to get out of their own way.

As such, I began to think about trying to find another framework to assess AI companies, which I’ll be diving into next week.

Read the original post here.