“Move fast. Break things.” How fast can we go, and how much can we break?
Investigating the limits of technological progress
Things are heating up
An interesting facet of technological and scientific evolution is that it is always accelerating. As an ever-increasing number of people “stand upon the shoulders of giants” and focus their efforts on pushing the boundaries, this acceleration is to be expected.
As this technology progresses, a question continues to simmer just beneath all the hype about innovation –what will happen to human jobs in the future?
The question is usually dismissed as unnecessary panic by the tech-optimists. And their favourite tactic is to remind us of how humans have been able to adapt to technological paradigm shifts all the countless times in the past.
Unlike disclaimers for investment products, techo-optimists seem convinced that past performance is exceptionally indicative of future performance.
But is that faith worth questioning?
I believe it is.
Questioning the techno-optimists
Techno-optimists seem to be unshakeable in their belief that human creativity will lead to newer & different types of jobs as new technologies emerge.
That the human world will adjust and chug along as it always has — similar to what happened when we moved from hunter gatherer to agriculture, or more recently the move away from manufacturing-based economies to service-based ones.
Unfortunately, I see a glaring blind spot — this logic only holds true when there is a large time-lag between creation of a new job category & the time it takes technology to do it better than humans.
The manufacturing to service-led economy made a lot of sense because services were something machines could not disrupt for decades.
But in our world, AI disruption is happening faster than ever — often in even less than a decade. As an example, I recall the software programming hype that started in the early 2010s. With “software eating the world”, it made sense to tell your children to prepare for a future with software programming at its core. Hopeful parents in places like India even took loans to give their children the opportunity to learn programming through start-ups catering to their FOMO.
And this decision seemed justified if one were to observe the sheer number of tech-related positions that college graduates would go on to fill in the last decade.
Perhaps a hope, albeit a transient one, was that programming could be the anchor to support the new wave of transition (similar to the earlier transition from manufacturing to services).
Fast forward to 2023 & we have very early versions of language models such as ChatGPT passing technical interviews at Google. I can only imagine these programs will get better at programming in the next ten years.
So the time lag between new jobs and automation shrunk from multiple decades to simply ten years. I suspect that even the ten years that it took in this case will be viewed as a “long catch-up time” in the future.
It is the nature of constant acceleration.
Furthermore, from creating art and music to playing chess, very few human endeavours remain as bastions of human dominance versus machines. And this too is likely to continue.
In order to find the role of humans in the economy of the future, we need to better understand the drivers of the time lags/ catch-up that determines the speed at which technology disrupts the status quo.
Drivers of the time lag (the past) — knowhow
In the past, scientific knowledge and technical knowhow seem to have been the primary drivers of this time lag.
Take the example of AI. The basic theory has been around since the 1950s — this includes complicated and modern-sounding concepts such as neural nets.
However, the computational power and methods to turn theory into reality was missing for decades. After many false starts, AI research truly began to take off in the last twenty years on the back of computing improvements and the explosion of training data that became available to researchers (I’m sure social media & apps in your phones had a lot to do with that).
However, it seems like this constraint will not be a permanent one.
As technology accelerates, we see the catch-up time between ideas and reality shrink. And this is likely to continue.
So is there anything else that could drive this catch-up time?
Drivers of the time lag (the future) — human pyschology
The idea that technology will continue to improve faster and faster seems obvious. With enough data and scientific progress, machines could eventually outperform humans at any task.
We already have tools like AI that can help machines do their jobs better and faster than humans.
But there are still some limitations. These have more to do with human psychology than with what is scientifically possible.
Let me explain further, but first we’ll have to make a detour and try to understand the difference between science & technology.
While science has generally been the purview of a few minds working on problems that may or may not have real-world application, technology is different.
Instead, it is like an offspring of science that is obsessed with real-world impact. It takes the principles of science and applies them in real-world setting to help humans. Usually by either de-risking our environment or making things easier (or more abundant).
This focus on changing the lives of an actual human being is one of the defining characteristics of technology vs science — even if we hardly think about it so directly.
The reason it bears articulation is that human beings have to accept technology or adapt themselves for it to have an impact. And since the impact or success of a technology is usually measured in the number of lives it manages to touch, adoption at scale is usually needed as well.
There in lie the constraints to technological dominance.
Constraint 1: Acceptance of tech
There are some areas of human activity that we may never be able to accept a machine’s output. Usually in areas which we perceive an emotional connection with another human being.
This is likely due to inherent empathy that humans are born with. We all have a deep desire to receive empathy, and even see ourselves in others.
It may be an irrational feeling, but it’s certainly present in our lives.
Any type of job that has elements of emotional connection is likely to be an area where human acceptance is going to be delayed — even if technology catches up.
As an example, think of our judicial processes. At this point, we likely have enough data points so that an AI could look at the precedents of a case and make a judgement in seconds. This judgement would likely be faster to get to (thus saving tax dollars), adjust for personal idiosyncrasies of specific judges and likely be fairer on average (esp considering we know human judges to make decisions based on things such as hunger).
But it is doubtful that humans would accept such a system. There’s something about making your case to a human being who has the ability to empathize with you that is irresistible. Even if we know that the average outcome could be less than ideal.
Similarly, psychologists or coaches provide more than just advice. They provide a feeling of being understood and heard by another human being. It is unlikely that technology can supplant these feelings anytime soon.
The other domain where acceptance will be hard to find are domains which allow us to day dream and take pride in being a human. Sports is a good example of this.
I’m sure it is possible to create robots that could shoot basketballs with an accuracy level that would embarrass Stephen Curry — but will anyone be interested in watching games where robots score hundreds of points against other robots?
At some point, the fascination of watching Stephen Curry is driven by this belief that he’s not different than us. He’s human too and has the constraints that we have. It is therefore a pleasure to see how a human overcomes those constraints. The fact that Stephen Curry and you don’t share the same bodytype or genes is something we’re willing to overlook, but it will be much more difficult to overlook the differences between us and a machine.
Constraint 2: Adaptation of self
Even in domains where acceptance of technology doesn’t represent an attack on our very humanity, there may be other constraints.
Although smart people can advance science for real-world use (i.e. technology) without limit, the average person can only adapt to change so much.
Change may be the only constant, but that doesn’t mean we love it. Instead, change is always a bit uncomfortable and many of us avoid it till we have clear indications that the change is indeed very beneficial. And to change too regularly, even if helpful, goes against our genetic instincts.
As an example, think about your phone. Suppose you rent your existing phone at a fixed rate of $30/mo. Now, how often would you upgrade your phone — assuming your rent continued to be $30.
If humans didn’t find change exhausting, the answer should have been “I’ll change my phone every time a better phone is released”.
If phone companies kept working really hard and releasing a new, better phone every single day, you would have to get a new phone every day. Most of us would find this prospect daunting. Possibly because something about constant change unsettles us — we have evolved to crave stability.
Human fatigue to constant change is likely going to be another important constraint in understanding how fast technology can disrupt newly created fields.
The economy of the future
We are at a unique place in human history. A critical junction where the rate of progress of technology will be driven by psychological human factors versus technical and scientific knowhow.
I suspect that the economy of the future will be split into three categories based on how psychology and technology interact:
- Hard to accept economy: Technology will continue to play a supporting character role in fields where we require high emotional connect to another human being.
As an example, AI could likely make the psychologist much better at his/her job, but is likely to take over the role of a psychologist completely. Even if it could have done a better job than a psychologist.
The timelag here is really high & human jobs are probably secure.
2. Easy to accept but high adaptation requirements economy: This category will be likely be related to businesses which depend on consumer adoption for their success (e.g., B2C business or businesses with a strong user/consumer use focus).
It will be exhausting to expect consumers to continuously change their mind on a given topic. Therefore while technical capabilities may race ahead, there will be limited demand for better products until a waiting period threshold is crossed.
I suspect that the waiting period would be different for different things. I.e., the time between regular upgrades may differ between a phone, a car and an operating system. More research into this “time-lag of adoption” would be interesting to see.
Human jobs are at risk, but not as quickly as your worst fears may lead you to believe.
3. Easy to accept and low adaptation requirements economy: This category will likely be related to businesses where mass human adoption is not a requirement. Instead, it would incorporate businesses where a small number of people are constantly looking for an edge (trading) or even specific processes in other businesses where machines interact with each other.
This will truly be the domain of AI’s constant upward progress, with humans being outclassed by machines in a very quick period of time.
Apart from the fact that we don’t really know the sizes of the categories above, a scary prospect is that technology will continue to make providers of services and products more productive — even in categories 1 & 2.
This means that even psychologists or jurists are likely to do more with less.
Unless we expect the size of the overall pie to increase in line with productivity gains, it would also mean that there will be job losses for humans. For example, if there are a hundred patients & a psychologist becomes 10 times as productive, you will need 1/10th the number of psychologists to serve the same patients.
The only way you could accommodate all of the psychologists’ jobs are by increasing the number of patients to a thousand. It is not at all obvious how sizes of pies increase vis a vis productivity increase. For example, the number of patients requiring mental health help may not be dependent on number of mental health providers.
In the event that the size of pie does not increase we will look at job losses. And unless these lost jobs are replaced by creating new jobs specifically in category 1, there will be a trend towards long-term unemployment for a large section of society.
Unlike techno-optimists I believe this reality is likely to occur. And an honest look into alternative systems of living, of giving meaning to human lives other than work will continue to become very important. And yes — I think we will need Universal Basic Income.
This is all great, but shouldn’t we be worrying about machines taking over and becoming our masters as well?
Not really. But explore that thought deeply here.
