What I Research

Human creativity is seemingly endless.

First, we invented agriculture. That let us settle down in one place and have surplus food. That meant some humans could do other things with their time, like invent tools or have discussions or build buildings. That was civilization. Then 11,800 years later we invented machines to do jobs for us: the spinning jenny, the engine, the car. Even $154,000 ramen robots:

 A newspiece I saw recently in China

A newspiece I saw recently in China

But then we did something different: we invented the computer, nanomaterials, and the internet. We used these tools not only to make our physical tasks faster, but to do the sort of things humans are really good at: judging, deciding, reasoning, creating. Information tasks. And so we used them to invent new things, like the “Three Horsemen of the Autocalypse”: self-driving cars, 3D printing, and genetic engineering.

This is our information age: technology that can think, plan, and do.

What Does This Mean for the Economy?

I’m interested in what this means for the economy. That’s because things like this have happened several times before. When they have happened, they’ve been the most important things that have ever shaped our standards of living.

First, we had the industrial revolution. Cotton-spinners were terrified by the spinning jenny – “if a machine spins yarn 100x faster than I do, then I will have no job!” (see: Luddite Fallacy) But over time, manufacturing technology became more productive and allowed more people to have better lives. Technology ushered in an era of prosperity:

"The most important graph ever"

Then the US had the Green Revolution. From the 1930s onwards, humans invented super efficient agricultural tools. Crop yields went up, food production went way up, and former farmers started moving to cities instead. (The share of US employment taken by agriculture went from ~40% in 1900 to ~1.5% today.) So we created cities that became more comfortable to live in, with all sorts of education opportunities and services and industries.

Today, we have intelligent automation. Intelligent automation does things like gives you better health diagnoses, spots potential school drop-outs, identifies faults in pipelines, predicts the weather, and allows factories to produce three times more for one tenth of the cost.

But it will also put people out of work. It will make a few people very (unequally) rich. It will require governments to rethink how it educates its children, invests in its industries, and makes economic decisions.

This is what I like researching: what are its risks?

This time is different

Why should we care this time? Every other time, we’ve gained unfathomably better living standards. Should this not do the same?

This is big because this time is different than every other time. Every time before, we’ve automated physical processes, like manufacturing sedans or fertilizing a corn plantation. Now, we can automate cognitive tasks and tricky physical tasks.

Automated production line at the Fremont Tesla factory

The Tesla factory in Fremont, USA. I once visited and was struck by the depth of automation. It turns out it's world-leading.

So what if automation becomes better than humans at some very “human” jobs? It’s possible that a big chunk of what CEOs, lawyers, consultants, chemists, politicians, and more do will be automated. This is going to put people out of work who may not have the skills to find non-automated jobs. This will likely increase inequality because the people who automate away tasks will either make more money from it or save more money.

This is a big, big opportunity

But this is overly pessimistic.

It’s not just that we create machines and wipe out human jobs in the process. The creation effect is just as important as the substitution effect. For example, entirely new areas have opened in biology for the first time. We can discover drugs using code and we can process human genomes for cheap. Biologists spend less time doing manual processing and more time doing the important stuff: analyzing results and thinking of new avenues. So we will need more biologists, and more new jobs we hadn’t even thought of before will open.

 A modern biologist

A modern biologist: 50% microscope, 50% computer scientist.

My specific research

Here’s where I come in. We haven’t yet studied this much in a quantitative and rigorous way. And we haven’t accounted for changes in the past 5 or so years. So we have little idea of what jobs, and why, will be automated.

Can we look at a job and figure out a) how likely it’ll be automated and b) for what reason?

Right now, I’m looking at an even more specific question:

Can we use the set of skills of a job and what the day-to-day job looks like to figure out how likely it is that job will be automated?

If we can figure these out, we could potentially:

  • Help governments set better economic policy
  • Help people choose jobs and education
  • Prevent social collapse from mass-unemployment
  • Identify new business opportunities

The tools I use

My PhD is in Engineering Science. People I talk to usually tilt their heads, squint their eyes, and say: “that doesn’t make sense…”

It does. I use “Information Engineering” to map, quantify, and understand these economic changes. A lot of this area is done in popular science and sometimes just speculation. I focus on using rigorous mathematical and computational tools to figure out these problems, because I’m a believer in good economic imagination and really good computational tools.

This will, necessarily, be a question of data analysis and computational statistics:

  • Lots of data: We have thousands or millions of characteristics about jobs and the economy for millions of people over dozens of years
  • Patterns in data: I believe that with more data there are predictable patterns that conventional tools don’t pick out
  • Sophisticated tools: I am trying to exploit the recent advances in computer science to make advances in economics – something that unfortunately doesn’t happen in a world stuck in linear regressions.

There’s a particularly beautiful tool I’m trying to use: Gaussian Processes. It’s a heavy focus of Mike Osborne’s group at the Oxford Machine Learning Research Group.

Where to learn more

First, start by checking out my exceptional advisors: Mike Osborne and Carl Benedikt Frey.

Then, check out this paper they wrote that caused an uproar.

Then, read any number of articles written lately that give a very high-level overview of the economic impacts of automation. [1] [2] [3] [4] [5]

Then, read any number of journal articles and reports. [1]

Then, read books: The Second Machine Age, The Future of the Professions (written by two Oxford researchers), Average is Over, Capital in the 21st Century.

“There can be no gainsaying of the fact that a great revolution is taking place in the world today… a technological revolution, with the impact of automation and cybernation… And there is still the voice crying through the vista of time saying, “Behold, I make all things new; former things are passed away.” - Martin Luther King