This page contains Chapter 10 from
Sustainable Superabundance: A universal transhumanist manifesto for the 2020s and beyond
Note: The text of this chapter of the Manifesto is draft and is presently undergoing regular revision.
To make comments or suggest changes in the text, please use this shared document.
10. Abundant creativity
[ This chapter is significantly incomplete ]
As discussed in the previous chapter, greater intelligence and more pervasive automation have the potential for catastrophic malfunctioning. After an initial period of apparently smart decisions and strong improvements in operational efficiency, these systems could veer badly off course, perhaps triggering an accidental nuclear war, or a meltdown in our global electronics infrastructure. This existential risk needs wise management, via the rapid development and enforcement of lean safety frameworks.
However, the concerns raised by AI go beyond the possibility of “killer robots” – automated systems that unintentionally kill huge numbers of of people. The concerns also include the possibility of “job killing robots” – automation that does workforce tasks much better than humans, and deprives humans of employment.
Accordingly, this chapter looks at the threat posed by greater intelligence and more pervasive automation to options for human employment.
However, rather than fearing this latter outcome, transhumanists look forward to the greater freedom that it can entail – greater opportunities for human flourishing. Humans will no longer need to invest such large portions of their time in occupations that are back-breaking or soul-destroying. We’ll be able, instead, to participate in the creation and discovery of music, arts, sports, ecosystems, planets, and whole new universes.
But before this abundance of creativity can be attained, some significant adjustments are needed in the human condition – changes in mindset, and changes in our collective social contract.
A short history of automation
Robots have been killing jobs, on noteworthy scale, since the first Industrial Revolution. Weaving machines were invented that could automate many of the tasks in the textile industry better than human weavers. Machines that drilled, hoed, rotated, or reaped dramatically changed the work of agricultural labourers. Assembly-line machinery transformed the work that needed to be done in factories. Word processors and spreadsheets – robots of a different kind – reduced the need for manual clerical staff. And that’s just the start.
A key response to various jobs being automated has been for people to learn new skills and change their occupation. Human labourers who used to work on farms moved to cities to find employment in factories, shops, restaurants, hairdressers, banks, and so on. Individuals have been able to use their brainpower to learn all kinds of new skills, enabling them to work in careers that could hardly have been imagined in earlier generations – including occupations such as software engineer, website designer, and social media coordinator. Other occupations, such as writer, teacher, farmer, and soldier, remain in existence, but are much transformed from earlier times; many tasks that previously consumed a lot of effort from the professional are handled nowadays by tools.
Will that same pattern be continued into the future? As robots and software improve in performance, will humans continue to be able to find new jobs for themselves, to replace jobs that have been made redundant due to the latest waves of automation?
That assumption can be called “the business-as-usual extrapolation” – the view that the future will remain broadly similar to the past. However, transhumanists recognise that the near future could see changes that aren’t just minor variations on the past; instead, technology has the potential to change matters fundamentally. In that case, extrapolating from the past into the future is subject to being undermined by major disruptions ahead.
Limitations to business-as-usual
The business-as-usual extrapolation assumes the existence of “uniquely human talents” which can continue to give humans a competitive edge in the employment market place over whatever automation can accomplish. In that case, so long as we humans are willing to be adaptable and to retrain, we’ll keep one step ahead of the robots.
Consider traits such as creativity, compassion, emotional awareness, personal coaching, concept formation, “common sense”, and intuition. These human traits may appear to involve features beyond mechanical computation. And consider the kinds of ad-hoc skills needed by a professional such as a plumber, for whom each new plumbing repair task might involve an unpredictably different configuration of pipes, valves, cupboards, and household goods cluttering up these cupboards. What kind of robot could deal with all that variety?
Transhumanists respond that there is no reason to believe in “uniquely human talents” that are somehow forever beyond the ability of computers to duplicate.
Indeed, impressions can be deceptive. Just because software, today, cannot perform a particular task, it does not mean the task will forever remain outside the reach of software. There has been a long history of tasks which initially appeared to be fundamentally beyond the capability of automation, but which were subsequently demonstrated as within the power of automation after all. One famous example is the task of driving a car. Arguments used to be given, that driving a car was inherently too difficult for any software to accomplish. However, over the last fifteen years, enormous strides have been made in the ability of self-driving cars. The question of self-driving cars has moved from an “if” to a “when”.
What lies behind much of these improvements are advances in computer hardware: faster processor clock-speeds, larger memory storage, and smarter, more numerous sensors. Even more significant is the enhancements in the software discipline known as “machine learning”, covered in the previous chapter.
Limitations to retraining
The business-as-usual extrapolation urges members of society to prepare to retrain more fully and more often than in the past. Such retraining may take some time, resulting, perhaps, in a temporary reduction in earnings. But with sufficient advance warning, employees can be encouraged to acquire new skills in parallel with still working on their old job. In principle, this will minimise the disruption they will face.
However, note that machine learning is a general purpose utility. Any improvements to the mechanisms for machine learning are applicable, not just to a single occupation, but to multiple different occupations.
Therefore, as robots are becoming capable of doing key tasks for Profession A, the same breakthroughs mean that robots are also becoming capable of doing key tasks for Professions B, C, D, and E. Truck drivers who lose their jobs because of improvements in self-driving vehicles may find that, by the time they have retrained to a new profession, robots can do that profession better than them as well.
Computer vision, which makes powerful use of machine learning, is one example of a general purpose skill. The same core skill that allows self-driving cars to reliably recognise objects crossing their paths will also allow workplace robots to reliably recognise objects passing through their environment.
The skill known as “common sense” falls into the same category. Common sense depends upon a large network of knowledge about real-world objects, including an understanding of humans and their motivations. Present-day chatbots, notoriously, display a low level of common sense. It’s easy to catch them out. However, it is only a matter of time before they improve to match the human level of common sense.
Even we cannot be sure of the timescales, transhumanists can make the following prediction: it’s going to become increasingly hard for humans who are displaced from one job by automation, to quickly acquire new skills that will allow them to carry out a different job that has no short-term threat of also being automated.
Limitations to robot-human partnerships
Robots have the potential to operate reliably, without getting tired, inebriated, distracted, or annoyed. They can communicate their learnings to each other, via “the Internet of Robots”, with the result that they can all benefit from the new experiences and insights of any one of them. Add in fast-increasing expertise in soft skills such as emotional intelligence, to their powerful computational skills and robust mechanical strengths, and robots become very attractive as replacements for temperamental human employees. Employers concerned about costs and about quality are bound to consider hiring fewer humans and more robots.
Defenders of the business-as-usual extrapolation acknowledge that automation will grow in prominence in the workforce, enabling cost savings and therefore higher profits. The defenders of this extrapolation suggest, nevertheless, that one result of these larger profits is that humans will be able to work in partnership with the robots that are introduced.
When costs reduce for part of a task (due to automation), companies can provide their products and services more cheaply, reaching lower price points than before. With larger sales volumes, overall profits can end up higher (even if unit sales prices are lower). In principle, companies can take advantage of these increased profits to hire a larger number of human employees. These humans won’t be doing the same tasks as the robots, but will be spending more of their time on the 20% (say) of their original job specifications which cannot be automated. These humans could also get involved doing new types of task that add even more value. In short, automation that destroys some jobs could actually create more jobs overall.
However, human partnerships with robots in the workplace are likely to pass through two phases. Initially, the combination results in productivity savings which allows business growth that in turn provides extra opportunities, overall, for the humans in the partnership. But in due course, the remaining tasks that the humans were performing will fall under the reach of improved robots, so that the opportunities for these humans in that workforce decline again.
As robots improve their general purpose skills, the second of these phases is likely to dominate the overall story. We humans will find fewer jobs available to us. Our contributions to robot-human workplace partnerships will diminish and diminish.
Unemployment and underemployment
The trends described above can result in two kinds of consequence for employees. In the first case, employees are made redundant by robots, and cannot find any good new job. As a result they become unemployed – experiencing “technological unemployment”. In the second case, they are able to find a new job, to earn some money, but that job falls below their previous expectations in terms of work satisfaction, intrinsic interest, and income level. In this case, the employee experiences, not technological unemployment, but technological underemployment.
Underemployment may be expected, when someone moves from being a comparative expert in one occupation, to being a relative beginner in a new occupation. What’s more, jobs that are low-paid give employers less incentive to introduce automation in order to eliminate salaries. So it’s no surprise if more and more people find themselves in jobs with comparatively poor pay.
This trend towards technological underemployment dovetails with other trends that have the same consequence of greater inequality in pay. These trends can be called “winner takes all”.
[More material is being added here]
<< Previous chapter << ===== >> Next chapter >>