There’s been a pretty good discussion on the BVN Slack channel, #future-of-work about robot overlords and when they are coming to take our jobs.

There was fair bit of discussion that’s been of the sort where we can’t seem to agree on basic premises. Once we got that sorted out things all seemed to flow quite nicely. I thought I’d write down some of those premises so that new people to the discussion can catch up quickly. Also to sort them out in my own mind.

What’s the point of this post?

There is going to be a time period that the future is going to unfold over. This seems to be a trivially obvious thing to say, but it seemed like the key to making everyone agree. I think that there are several periods of the past and the future that are useful to think about.

The point of this post is to outline what I think those periods are. This also makes what I think explicit and makes it possible for others to agree or disagree with me. Once we know what we’re disagreeing about it makes it much faster to come to something that we do agree on.

The eras!

State of nature

This refers to the pre-Talorian period, maybe even to a pre-Adam Smith period, where people just made what they needed. Maybe they went to someone else and bought a thing from them but that thing had been made from scratch by that person.

Industrial

In this period pin makers are breaking up their tasks, steam mills are making weavers much faster and production lines are banging out Model T Fords at a rate of knots. Technology is starting to put people out of work. That technology could be a machine (steam mills), the application of an idea (specialisation), or the combination of the two (production lines). In this period human manual labour is being replaced by the brute force of machines. Because we needed fewer hands to carry and bash things people were freed up to do other things, like become travel agents or to sell cars.

Information

This is the period that we’re coming to the end of at the moment. In the industrial period the machines amplified our physical capacity by doing the grunt work. They didn’t do any crafting they were replicators. People did all the thinking. In the information period people still do the thinking, but the grunt work is done by machines. E.g. people put information in once, and then it gets fed to different places. In the industrial era that would have meant filling in forms in triplicate etc.

This has led to the travel agents having a tough time. This isn’t because the computers are doing anything special, they are just better at searching a lot of databases very quickly. There is a bit of a lessening of the service provided by a good, human travel agent because they can’t anticipate what you want, or offer useful alternatives. However, computer systems are cheaper than people. We book through a computerised system because we value the extra cash in our wallets over the better service offered by a person.

Species changeover

we’re coming out of the AI winter

This is a controversial name, so if anyone has a better one I’m all ears. The increasing ability of computers to do things that are unique to human abilities has been underlined recently with the defeat of a Go master by a computer. This seems to suggest that we’re coming out of the AI winter. This is not the part of human history where we are replaced by Skynet! This is the transition period. It’s just starting in a serious way, and it’s going to be a really interesting time to be alive¹.

This is the bit I’m going to spend the most time writing about because the next period looks pretty grim².

While we navigate this period (of unknown and probably unknowable length) there will be a few important things to think about.

1. Humans have money. At the moment no machines have agency (that we know of). Even the high frequency trading bots are explicitly doing the bidding of a human somewhere.

2. Humans care about ‘irrational’ things. We care about the story that a barista tells us about a coffee (“oh, I can really taste the life story of the grower” vs “ooh, this is black and hot”). Some of these things can be made to seem rational by casting them as signalling, but we certainly don’t think of them that way. Because of this humans care about doing things that aren’t well aligned with computer’s wants.

3. Humans are good at human things, computers are good at computer things. That is a well understood distinction, but recently the line between the two has been moving over towards the computer. Computers can now recognise and describe objects in images, they can drive cars, they can win at Go. They are pretty terrible at opening doors and having goals of their own.

   Embedly

4. OK, that’s the good things out of the way. Thing four is more or less thing three, but with a pessimistic slant on it. The things that you value as your key, human, irreplaceable traits are probably next on the list for AI to conquer. Things like coordination and negotiation are pretty human skills, but they are only really needed in situations with incomplete data.

   coordination and negotiation are … only really needed in situations with incomplete data

   For example, in a construction context, architects talking to builders is largely to discuss things that were missing or unclear on the documents. This isn’t a bad thing now because there are gains to be made in intuition. In the future I’d imagine that the documentation will me more complete, and will contain conditional information³.

   It’s important to remember that this period of history is characterised by a new type of automation. Previous periods have been about unthinking work, moving documents and information down predefined paths, lifting hammers that are too big for a human to lift etc. This era, the era of machine learning, is about automating thinking tasks. We like to think that our key value is our ability to think, but actually, our ability to work in confined spaces might be just as important⁴!

5. Most of what you do with your life is not high value. I feel like I’m a smart valuable person. You probably do too. However I generate most of my value to an endeavour in very short bursts. It could be 10 minutes of work that turn into a month’s worth of my output. Some people are able to have higher outputs, but they are probably also assisted substantially by a support structure. Most of my life is spent getting ready for the next burst (reading, resting, doing important, but low value tasks). Hold that in your mind for a second.

   Part two of this is that it’s very unlikely that a machine will “take your job” in the sense that they will replace you 100%. What is far more likely is that the machine will erode your usefulness. If you are a thing-doer and your company employs 100 thing-doers, then it usually won’t be a situation where they call you all into a meeting and say “someone has automated your jobs away, goodbye”. Your job will become a bit more efficient, then when Bob retires and Mavis moves to Manitoba they won’t be replaced. Then 98 of you will do 100 people’s work and it won’t be a stretch.

   work will be hollowed out of all the bullshit that can be automated

   Putting these two together, what this will mean is that the effect that has hollowed out the demand for village musicians will do the same for most knowledge workers⁵. The work will be hollowed out of all the bullshit that can be automated, and therefore will require fewer people to do it. I’ve already talked about delegation and this will be another example of delegation to a non-human agent. The amount that this is possible will vary by occupation, so to get a good idea of how hard yours will be take an inventory of your week using Rescue Time or something to give you an accurate picture of what you actually do. Then decide if you are the only person who could do it. The percentage of stuff that only you can do is probably the percentage of people who will be left.

   The other part of this is that your job isn’t a platonic whole. It’s a collection of tasks that cohere together in some sort of more or less meaningful way. As well as whittling away at your usefulness by some process of attrition, by knocking a percent or two off your to-do list, there’s also a chance that you could just be restructured out of existence. One process could take 1/3 of your job, another could take 1/4 and the rest could be spread out over your colleagues. These are the big layoffs we see at car plants.

   These two things are just opposite ends of the same spectrum, but they look different to us. The latter is much more dramatic, but on aggregate has the same effect.

6. I’ll leave this list on an optimistic note. All of the above doom-mongering is based on a simple (and wrong) assumption that there will be no new jobs created. There will be growth in sectors that are currently small, and whole new sectors created. To take the shine off the enthusiasm, there’s no guarantee that new job creation will continue to outpace deconstruction. This might be the time that things get out of hand!

The question for this period are how best to navigate it so that you hang onto the mechanical bull of the high life for as long as possible. Eventually it’ll be about owning capital; especially land or raw materials⁶. In the transition it will be different. I can only guess at how, but I’d say that it’s about meta-skills. Can you be really charming? Can you learn new things really fast? Can you predict where things are going and position yourself to catch the wave rather than get pounded into the sand by it?

Homo-redundant

By this point AI has surpassed all of humanity’s capabilities and is now the dominant species on the planet. There are a bunch of different ways this could go down. For a fairly gentle explanation, wait but why is excellent and for a really kick ass explanation, go straight to the source with Nick Bostrom’s book.

Should I be really pessimistic?

I don’t think so! I think that the way to go is to be an informed optimist. There might well be a change-over period longer than all of our lifespans, so going off grid right now would most likely be foolish.

Working on your meta skills, and finding pockets of new types of work will keep you ahead of the automation. Learning to work with computers and robots will make you safer too. I’d suggest inviting them into your life to free you up to do other things, things that you are more uniquely suited to as a human.

“References”

Here’s some of the links that have been posted during our discussions:

The Deactivation of the American Worker

The Post-Cubicle Office and Its Discontents

This book is short and gets referenced a lot. If you haven’t read much about this subject then it’s a great starter. If you have, it might be a bit basic. Maybe their next book is more up to date. Race Against The Machine: How the Digital Revolution is Accelerating Innovation, Driving Productivity, and Irreversibly Transforming Employment and the Economy

I’m listening to this as an audio book at the moment. It suffers a bit from the same problems as Race Against the Machine in that it seems a bit basic/obvious. The Future of Work: Attract New Talent, Build Better Leaders, and Create a Competitive Organization

Practical ethics is a great blog. This post, A jobless world—dystopia or utopia? has a good conclusion:

the challenge is to prepare for the coming of the machines with policies that diminish extreme inequality

that fits pretty well with how I see the challenge of this era-shift. A jobless world—dystopia or utopia?

Things computers are good at: Google’s DeepMind defeats legendary Go player Lee Se-dol

Another Win! Google’s Artificial Intelligence Beats the World Champion of Go For the Second Time

Things that they are not good at: Embedly

This shows that computers can do tasks that we think of as uniquely ‘creative’. The computer hasn’t done the composition, but it understands enough of what is wanted to do a very good job of filling in the blanks to complete an intent. This is what I mean my support or delegation. neural-doodle

Great article about the implications of the structural change caused by mass unemployment, and where that unemployment is going to come from. Self-Driving Trucks Are Going to Hit Us Like a Human-Driven Truck - Basic income

New jobs, but will their creation be fast enough to keep up with the losses? Will the losses be in groups of people who can do these jobs? Maybe we’ll end up with not enough people to do the new jobs and too many people who did the old jobs until they disapeared? The Most Important Design Jobs Of The Future

These Will Be the Most Popular Jobs in the Future

Our tech future: the rich own the robots while the poor have ‘job mortgages’

This is a gateway to a whole special edition about the future of work. What Google Learned From Its Quest to Build the Perfect Team

A dry and nerdy slide deck about AI startups. The State of the AI market in 2015: A focus on exits

These two links were from a discussion about what it would mean for architects to with with a continuous integration server like a lot of programmers already do. There were, of course, the standard objections that it’s not possible. I think that is a really dangerous attitude, the opposite of what I mean by inviting the robots in. To survive, businesses need to take every possible bit of automation they can, even automating things that they don’t do at the moment because those things are too much effort. Autodesk Revit Model Review

Owner Standards Check

How much should we fear the rise of artificial intelligence?

I can’t recommend this book enough. You can read quite a lot just on the website, but it’s a fascinating book. I find myself going back to it again and again. A History of the Future in 100 Objects

This is a graphicalisation of the Michael Osborne and Carl Frey paper about how easily different jobs can be replaced by automation. Machine Learning: the future of employment

Most of us have little empathy with finance people, this might be easier to read in a dispassionate way, and then to try to apply it to your own job. The Robots Are Coming for Wall Street