Work In The Face of Automation
A possibly nauseating diatribe/treatise on predictive models and "The Future of Work"
I was a part of a bioinformatics research lab while in undergrad. I did work in what was a tiny cubicle with the occasional all-nighter studying for various exams. The work that I would do in-between assignments was menial and unfulfilling grunt work to support the doctoral researchers. I would perform tasks like cleaning data, converting scripts from Python 2 to 3, and spinning up data pipelines. It's a well-known joke in research circles that the quality of grad student code isn't quite the best, so much was done in the code archeology department. Throughout my semester's tour of duty, I mainly performed no work in what was then a core interest and mostly focused on the ternary operations to ensure that papers and drafts kept on time. Don't get me wrong; I was happy learning and making sure the lab's software and processes kept in check. However, the work was meaningful to me only because it had a positive effect on the team's morale. In the back of my mind- playing hard support maybe wasn't the idea I had when signing up to work on papers there.
My career eventually went into another nonacademic, personally satisfying, direction. Funnily enough, this experience was one of the first thoughts I had when witnessing the awe-inspiring brilliance of OpenAI's work. For those who aren't aware, OpenAI is an A.I. research and deployment company that just released GPT-3. GPT-3 is a model that aims to achieve language understanding using unsupervised learning methods. There's a lot of statistical learning terminology thrown here. Although this isn't a post explaining GPT-3, think of it as a system that can generate text on a subject provided, you give it an enormous volume of content that it can "learn" with. For more information, here is a link to a layman level article from the MIT Technology Review with an explanation and internet reactions. Now, at the time of writing, not everyone has access to the model. OpenAI requires an application and approval process to get access via their API. Still, examples in the wild show the ability to generate front-end code and write about niche topics like information suicide. I want to be clear, GPT-3 isn’t the super AI that would be the end of human work but it did bring up some conversations among my software developer/designer friends on if would they be okay in the coming years when junior devs might take their opportunities.
In the past, conversations around automation have centered around the effects on blue-collar workers. Although we aren't discussing the impact of macro trade on this post, there's a misconception that American factory workers are only losing their jobs due to international labor pressure. The other major one is automation, and after companies recover from economic shocks, they restore to full productivity with fewer employees. (Note the COVID impact)
As can be referenced in the above graph, post-NAFTA manufacturing productivity has remained somewhat constant but has an annual decline of .5 percent on it’s employment levels. And although indeed, the U.S. isn't growing its manufacturing sector, it has been able to supplant those people with new and more advanced automated systems. Although the real domain of what to do with disaffected Americans is a political problem as much as it is economical.
Much have lamented about the incoming demise of the web development profession, journalism, and any knowledge worker. Not to mention years of "right to work" legislation weakening U.S. worker protections that might contribute to such demise. I feel these fears might be overblown. The difference between the U.S.'s comical mismanagement of its blue-collar labor force and its knowledge worker compatriots is demand and ability for demand generation. The ability of blue collar workers to spin up their own business usually tend to be excluded in commercial lending sectors. Whereas the barriers to starting a business with knowledge work are significantly smaller. With an expected dissolution of the rough equivalent of data plumbing and simple graphic design engineering roles- I feel that advanced models after GPT-3 can serve as tools that will help the labor market transition to more meaningful work. Companies can move faster, possibly run leaner, but enable a new wave of new companies and more profound roles where people can focus on decision making and not on the execution phase.
Although there is an expected change to what careers might look like in 10 years. Maybe more so than the last 10 or 20 years combined. But I believe that these professions will be just fine as sign painters moved on to vinyl sign printing. We should be embracing these models for a future of assisted labor where this can open new opportunities for all. Non-technical political campaigns who starve for technical assistance can finally “speak” to existence a campaign website where they might have paid too low a wage to afford a web developer. Small industrial businesses can expand their operations with a better copy with the A.I. assisted copywriting tool where they might not have otherwise hired an advertising consultant. Web developers can focus on accessibility and more creative experiences beyond the standard bootstrap website. With that said, I am not delusionally optimistic and understand the implications of labor upheaval.
Initially, I am reminded of a possible parallel. The Enclosure Movement was when Britain privatized the commons and transitioned it toward single ownership models of land. Contrary to popular belief, this action is believed to be supported by the citizens. It ended the cycle of subsistence farming and the cycle of poverty that allowed Britain to take advantage of specialized farmers who know how to get the most of the land. With fewer subsistence farmers, the spare labor capacity help gives way to the industrial revolution. One clear difference then compared to now is that compared to then- many people feel like they are looking down from a cliff rather than looking to a quality of life improvement. The question is, what would spare labor capacity of knowledge workers look like? Initially, with the rise of flexible labor markets such as gig-work platforms, surely have applied negative wage pressure. Assuming said jobs disappear for good, where might these people go?
A possible answer can be gleamed from the securities clearing industry. You would think that walking through the near empty halls of the Chicago Board of Trade that the demand for equities has fallen to new lows. Gone are the days of roughneck traders standing in full trading pits. However, as everything moved online, the clearing industry has only expanded. It's trading volume and manage to employ and support more significant amounts of people. New brokerages and securities need servicing and despite the increasingly more stringent regulations that investors face. The financial industry (if you remove teller agents) is still growing at a sustainable pace where new opportunities lie as upstarts aim to increase financial access. As models come in to automate and improve accuracy, more complex financial instruments will arise. Even opportunists and grifters alike found a new home in cryptocurrencies where new business is thriving there. Besides, the engineers and program managers, along with the latest compliance folk alike, are supporting these new-found technologies. Similarly today, despite the enormous breath of western and eastern software giants- there remains significant market development in the software and hardware products of tomorrow.
Now assume if newer models and A.I.s do truly eventually effect work to the point we reach record levels of unemployment in the knowledge worker spaces. We should then approach the quantity of spare labor capacity, not with a cynical mindset. Instead, ask what incentive structures we can place in the market to decrease living costs to make it so that all than thrive in a world where we might face downward wage pressure. In addition, maybe with spare capacity we will head into investing effort into innovation capacity, finally being fulfilled. There is a debt in American infrastructure that needs to be paid. There's plenty of industries that could use a new perspective once the era of the bullshit job is behind us. It will require a level of cooperation that has not been imaginable in the last 20 years of history but such tantalizing opportunities can lead us to the next generation of technologies.
For the few software developers reading this, you know that at many big companies, you have large, ambitious projects that have been rotting in Jira because the effort was too significant to make those bets. As long as you are committed to seeking new problems and solving them- you and the world will be fine. I think back to the first few roles I had as a junior developer and even the menial tasks that I have done. I don't know if the lack of entry-level opportunities in specific areas will stifle the profession, but in many ways, I wish I was able to cut to the chase on the critical goals I needed to do. Maybe focus more on designing resilient systems rather than writing tests. Although I know GPT-3 is essentially a more powerful Google search at the moment, it would be nothing short of awesome to have more fruitful suggested replies to emails so I can do other things that matter like being with the ones I love. Maybe I am just wistful for a new technology to bring out meaningful societal progression.
I want to close with this- because we do face a significant challenge on how work provides meaning. If we are to move on from the jobs we have today to the ones we need to do tomorrow, we need to address how difficult it is to build physical infrastructure. We also need to address the western economic bifurcation egged on from dangerous policy decisions. Many of those with access to capital have benefited most from the productivity gains of their money. For defenders of capitalism like myself, in a world where labor looks like a cost center needing to be cut in a company's balance sheet, a real question must be posed to our governments. What is governance's role when markets might not immediately know what to do yet with its participants?
Angelo Saraceno is a Product Manager at an enterprise SaaS company. In past roles, he was an intern at Adobe, at a hedge-fund and dabbles in entrepreneurship. He is always availible to talk about nearly anything interesting, you can reach him on https://twitter.com/ndneighbor