It is almost 15 years ago, when I first had a touchpoint with the ethic and morale question: “How human is automation after all?” At the time, I was writing software and automated the process of mail and parcel sorting for the major postal service in Brazil. So, I was at the construction site in Rio de Janeiro, commissioning our software that was a central part to a new parcel sorting system. This system could sort parcels with superior efficiency and speed, replacing the workforce of about 30 human workers.
In the manual process, people picked a parcel from a belt, scanned a barcode and got a destination information from a computer that was part of their workstation. Based on the result, they tossed the parcel onto another belt, a bag or cabinet. The new automated process solved this by fully automated systems, using automation equipment and software. If you like to get an idea how such systems look like, here is a promotion video from Siemens.
During a lunch break, I happed to have a casual talk with one of the workers, based on the great topic of football (soccer) and how German and Brazil talents may compete. Then he asked me the question: “Why do you actually participate in the destruction of our jobs by installing this system, you seem to be a nice guy personally?” Think of that. He figured that I am a nice person personally, but at the same time are one of the devils who kill his job and so put his life-basis at risk. 15 years back, there was no backup system for people who lost their job. Most probably, their entire family would massively suffer from poverty immediately. This caused trouble in my head, making me struggle with my own value system and passions, for a long time.
The automation in production, manufacturing, logistics and many other domains has killed many jobs around repetitive, physical work. One study estimates that about 400.000 jobs were lost to automation in U.S. factories from 1990 to 2007. While this is a major political, economic and social challenge since many years, the individual fate for many people was fatal. At a global scale however, many new job roles were created out of it and the fear of massive, global unemployment at scale, because of automation technology, did not became reality. These new jobs however were on a more cognitive level, e.g. automation engineers, programmers, data analysts and these days machine learning and AI specialists.
What happens right now is the next level of this evolution, where machines (physically as robots and on cognitive level as algorithms) take over more and more cognitive jobs. Jobs that required pure physical skills are in danger for 2 decades now, but the new development is that jobs replace repetitive, cognitive jobs as well. Examples are: Translation of texts, screening of legal paperwork, calculation of your taxes, stock market trading and answering knowledge questions. Just think of the skills that systems like Amazon Alexa and Google Home bring to your home. They understand your questions and find the music you want, provide you with knowledge and can control your home automation equipment. There are thousands of companies who develop and invent technology that is designed to replace humans in cognitive tasks already. Yuval Noah Harari wrote an amazing analytical book around this topic (and more): Homo Deus – A Brief History of Tomorrow. If you want to get a very holistic impression, I can recommend this view point.
Here is a nice article in the time magazine, about how the Covid-19 pandemic boosted this process: “Millions of Americans Have Lost Jobs in the Pandemic—And Robots and AI Are Replacing Them Faster Than Ever”. Even more radical is the headline of an recent article in the Forbes magazine: “Artificial Intelligence And The End Of Work”. This article introduces with the following statements:
Stanford is hosting an event next month named “Intelligence Augmentation: AI Empowering People to Solve Global Challenges.” This title is telling and typical.
The notion that, at its best, AI will augment rather than replace humans has become a pervasive and influential narrative in the field of artificial intelligence today.
It is a reassuring narrative. Unfortunately, it is also deeply misguided. If we are to effectively prepare ourselves for the impact that AI will have on society in the coming years, it is important for us to be more clear-eyed on this issue.
So, the question that I was confronted with 15 years ago in Rio de Janeiro is more relevant than ever. During my research about how other people deal with the question I came across an interesting web-site: https://willrobotstakemyjob.com/
This site tries to calculate the risk for any job in their database, to be replaced by AI sooner or later. It also tries to give you advice, if you should consider changing career plans in accordance to their data. Now guess what powers this engine…? Data Analytics, algorithms and AI. Here is a computer system that gives humans advice how to deal with the fact that AI will replace them at work. Ironic. By the way, you do not need a human career consultant for that matter anymore, as it seems.
I think, ultimately, every repetitive process that can be modeled as a system of inputs and outputs and where all inputs and outputs can be represented by data, will be automated by machines. May it be physical or cognitive tasks. There are 3 major technical challenges for all these systems:
- Figure out how to represent all required inputs as data. For physical metrics we need to invent sensors and data models to represent them. We need to integrate data silos build new interfaces and harmonize data models so that a computing system can leverage them. Today many data points are not integrated yet and not all inputs can be represented as data. But this is changing quickly. Just remember since when it is possible to represent a complex traffic situation as data by using cameras, ultrasonic sensors and radar as sensing technology. Since when is it possible to sense human thoughts to control mechanical limbs? It is just a matter of time and invested energy until complete coverage of a physical metrics that can be used in automated systems. Data integration is an old topic and solutions become smarter, the same applies to harmonized data models.
- Figure out how to map all generated outputs to action. Computers, algorithms and automation equipment outputs data points. A control bit changes from 0 to 1, a machine learned algorithms predicts a failure at 95% or a gaming bot decides to move a virtual avatar. In all cases, the output is data, we just need to figure out how to turn this data (or insights or however you call it) into action. So we need data-to-reality actuators. In automation today, the solution is realized by equipment like PLCs with physical actors like motors and air-pressure valves. But also autonomous cars use motors to do steering, acceleration and breaking. We are just at the beginning of fully automated, so called closed loop, self-optimizing systems where a system measures, controls and optimizes without any human intervention. Until we fully trust the algorithms, the output of machines will mainly focus to inform humans so that they press some buttons, but more and more the output will be directly linked to physical and virtual actors.
- Build the algorithms. Here algorithms stand for any computer based input-to-output calculation. This can be designed behavior (deterministic, programmed behavior of computer programs), learned behavior (trained algorithms, e.g. neuronal networks, either supervised, unsupervised or re-enforced) or evolutionary optimized behavior (feedback-based selection of random mutations of machine behavior). I specifically think of the latter as the most powerful way, while humans and machines may tell a system if it worked well or not, it tries out slightly modified versions of itself in massive speed. The mutation that works best moves on. While nature needs millions of years to create complex systems out of natural evolution, humans may be able to build machines that can evolutionize at blazing speed within days.
Just think of the idea of quantum computing and how many parallel mutations of a certain algorithm can be computed in parallel to find the optimum… Hard to imagine? Considering the history of computing, it is realistic to assume that algorithms and robots will be much better than humans, at any task that repeats physical work or cognitive input-to-output work. The only question is: when will it happen? Some of that stuff is already happening, some is already commercial, some is more scientific and research.
So – back to my initial question: “How human is automation after all?” – and – “are you working as an enemy to society when you help customers to leverage latest technology to replace human labor (just to save money)?”
In my mind, it is in the nature of humans to become better. In nature, single pieces of matter are using evolution to build more and more complex structures. Dust condensed into planets, stars and galaxies. Organic cells organized into simple life-forms and used evolution to build complex organic machines like the homo sapiens. Now homo sapiens follows this nature to build more and more complex non-organic machines. It will not stop, for me it is a force of nature that humans will simply not stop to invent new technology that will ultimately outperform humans. Because the non-organic machines will use evolution cycles that are in the area of months, minutes and milli seconds. Companies and people who ignore this, will be the first to become irrelevant while those people who embrace the idea and invent this technology will be very powerful, mighty and influencial (and rich).
I personally want to understand what’s going on and I want to stay relevant. So, I decided a long time ago that I will not step away from building automated systems out of that reason alone. Moreover, I try to keep up with the lasted developments in AI, ML, IOT and quantum computing, becoming a part of the transition – not just an observer. Do I accept that people need to suffer from this development while others (much less in number) benefit from it? No! In my belief, we need to automate more, develop the technology further to leverage the benefits and bring humankind on a next level. And that level should be a place where humans can actually omit 40 hours work-weeks and focus on science, art and entertainment and live a life free from hunger and sickness. At the same time, we have the political, economic and social responsibility to also take care of the people who will actually loose their jobs in the next years and make sure that their suffering is limited or completely absorbed.
Specifically, the political influence is required to give purpose and income to people who do not have options in the future job market (not everybody can become an AI engineer or robot designer). At this point, I like to call out that everybody who is at the forefront of technological advancement as scientist, researcher, developer, architect, consultant, big data engineer or whatever your job role is, you need to also play your part to think about the effect on the people. Take influence in the political discussion and help to ensure that we all benefit from the technology, and not only those who invent and commercialize it just get insanly rich and powerfull.