Consumption Expansion (Edition 18)

'Consumption expansion' might be the most significant concept that you probably haven't heard of when it comes to AI. Especially in the context of jobs.

Note: This post focuses on a more generic AI topic, although one that has significant implications for Sales. Jobs, rather, job losses. So read a few lines and decide if you want to read further or not - after all it has a direct impact on each one of us.)'

(Yes, it’s slightly silly image. Just not worth spending more time.)

Putting the extreme bear case ‘AGI will wipe out all humanity’ scenario aside, ask anyone what’s their biggest concern about AI; and odds are that you will hear the word ‘jobs’.

A lot of people, including some really smart ones, seem extremely concerned about AI taking over jobs. And complete chaos - economic and social - ensuing.

People are worried that what they have painstakingly learnt over the years will suddenly become obsolete. They are worried that they will not have anything to do (or perhaps that they will have no way to earn). They are worried how they will support their families. They are worried that life as we know it might change irrevocably.

And they probably have a good reason. A lot of smart people have been sounding the alarms. Bill Gates - the man who brought us personal computing (sort of) - says that humans won’t be needed for most jobs in 10 years. Vinod Khosla, a contemporary of Bill Gates and another very learned man, says something similar about AI replacing jobs from farm to the office. Elon Musk has said it (although one could say that he gravitates towards the extreme bull case on everything related to his companies and extreme bear case on everything else) Bloomberg says that Wall Street is set to shed 200K jobs because of AI.

Is it all doom and gloom then? Is it all over for us humans?

Unlikely. In this post, I am going to help you understand why most of these smart people are extremely likely to be wrong!

---

The answer to that lies in a term that you have never heard of, but some economists are familiar with.

Consumption Expansion.

‘Consumption expansion’ is a simple term with a very literal meaning. It implies a scenario where ‘consumption’ a.k.a. usage of something increases and expands. Where demand for a certain good or service rises disproportionately to other goods or services as people’s awareness of that good’s (or service’s) utility rises. Where people start using more of something because it allows them to start doing not just what they could not do earlier, but even what they could not think of earlier.

Now this is something we all know and not exactly groundbreaking, you might say. And that wouldn’t exactly be wrong.

But there is one more aspect to expansion of consumption that most of us tend to miss. That people start doing more of the new thing, but they also keep doing the old thing. Sometimes even more so than earlier!

Let’s consider a few ‘things’.

For most of us, Mainframes are a thing of the past. A thing from the 60s and 70s. Would you be surprised if I told you that the Mainframes market is actually growing at a CAGR of 7.9% even right now?

Now you might say that Mainframes have specific utility in certain sectors. Then let me talk about something else - something that Gen Z and Gen Alpha might not even have heard of.

Fax.

Remember Fax? That thing that peaked in the 80s and went obsolete by 2000.

Or did it really?

Turns out that Fax usage is actually growing at 25% per year. In 2025!

Oh, if that doesn’t catch your attention, let me go back into the past a little bit further. Let’s talk about horses.

Yes, horses.

Do you know that the US today has more horses than 1960? What if I then told you that it now has more horses than even 1860! (I wrote about this earlier; incidentally in a post where I introduced the ‘expansion of consumption’ concept).

Now, this term, and its use for technology consumption might be new, but the concept has been around in Economics at least for some time. Consumption makes for a large part of GDP, and consumption led GDP expansions have been the norm in most high-growth economies. Or at least a significant presence.

‘Consumption Expansion’ is also the reason why factory jobs still exist. Or farm jobs. We consume so much more of everything today that even when percentages decline, the absolute numbers often do not. And sometimes even go up!

There are still more farm jobs in the US today than software engineering jobs. 2m+ compared to 1.6M+.

Good news is that not everyone in the world shares Bill, Vinod or Elon’s concerns. Popular economist Noah Smith has argued, quite eloquently, that jobs might be plentiful in the age of AI. Both more and better. Definitely better. Economist Tyler Cowen has said something to a similar effect.

So it seems that for once, the bean counters might know more than the geeks. (I see economists rising in unison “we are not bean counters!” Chill, it’s just for giggles, people.)

This is anyway a point of view that I definitely subscribe to.

There are more reasons why jobs are unlikely to go away. For e.g., when it comes to programming jobs, I have argued that LLMs are, ahem, ‘just’ another layer of abstraction. A paradigm shifting layer, which will democratize programming for billions, but a layer nevertheless. And the beauty is that there have been at least 5 layers of abstraction in the past. I would even argue that moving from physical punch cards to directly typing a program in the computer’s memory via a keyboard was as big if not bigger - because we managed to change ‘real world’ mediums. We moved from a physical interface to pretty much a virtual one!

Over 000s of years, there have been more revolutionary inventions. From the steam engine (industrial revolution) to agriculture (society as we know today) to fire (civilization as we know today). And I’d argue that some of them were ‘relatively’ bigger than the LLM architecture.

(Why ‘relatively’? Topic for another day.)

Now, this is not to say that no jobs ever go away. They do. Sometimes completely, sometimes partially. Sometimes slowly, sometimes quickly. We barely hunt anymore, we barely farm anymore. We will barely work in factories anymore.

When it comes to AI, I think this is one place where a Wall St. firm might have gotten it right - in 2023, Goldman Sachs estimated that ‘only’ 7% of the jobs will be substituted (a.k.a. replaced) by AI. Another 63% will be complemented, and 30% will remain unaffected.

The numbers shared by them are unlikely to hold. Fundamentally, however, I believe this is the right framework to work with - where a smaller percentage of the jobs will get completely eliminated, a large part will combine humans and AI, and a significant enough part might not see a large enough impact.

When this happens, it could turn into a period of temporary joblessness and high uncertainty. It will be temporary, but it could still be a long enough period.

---

Before I end, let me add that I do have a fairly big concern myself regarding AI’s impact on jobs.

I wonder what will happen when AI intersects with robotics.

I have always believed that white-collar professionals have significantly more ‘skill mobility’ than blue collar professionals. And if you are not aware, robotics is advancing rapidly. Some people expect that 2025 could be the year that robots start going mainstream - especially humanoid ones.

I have no doubt that the world will do just fine, like we have for 000s of years. However, it’s quite likely that the change will be more drastic. That there will be more turmoil, and more long-term effects.

And consequences. That we might or might not be ready for.

At least yet.

Challenge if someone is keen : Show me a LLM that could have written this post.