Let’s forget about taxes for a moment and refocus on the potential jobs that will be wiped off the map. To understand the gravity of the situation, I’d like to highlight the development of the AI software called Amelia, by the company IPsoft. This program initially analyzed and learned how to perform the work of call center employees, meaning telemarketing and customer service tasks over the phone, improving with each interaction. In one of its first pilots in 2016, this AI worked for a company handling about 65,000 customer calls per month. Just six months after the start, following several corrections, Amelia was already able to successfully resolve 64% of inquiries. Not only that, but this also reduced the average call duration from 18 minutes to 4.5 minutes and the response time from 55 seconds to just 2 seconds[47]. In the short term, this presents the potential risk of taking away the livelihoods of several million people worldwide. Today, Amelia provides solutions for Telefónica, as well as for banks like BBVA, insurance companies, and many other industries. In a similar vein, in July 2023, Summit Shah, CEO of Dukaan, laid off 90% of his staff due to the implementation of a bot similar to Amelia, developed by one of his employees in just two days[48]. After the bot’s implementation, Dukaan reduced the resolution time for each inquiry from 2 hours and 13 minutes to just 3 minutes and 12 seconds[49]. But the blow doesn’t end there. Every unemployed person is one less consumer of basic goods and services, from candy at a kiosk to transportation services, food outlets, and clothing, creating a domino effect in various sectors of the economy.
This same scenario can be replicated with similar technologies in other types of jobs, like cleaning, building a bridge, providing directions, or retail sales. And of course, the response of a techno-optimist is that this actually frees humans from repetitive and tedious tasks so they can perform more complex ones. And yes, for some workers that will be the case, but not everyone will be so lucky. Why have a hundred people working on a customer service line if an AI can handle most inquiries on its own and only escalate those beyond its current capabilities to a small group of humans?
Many raise their voices to say that it’s false that AI will take jobs away from people, and that it will actually be a person who knows how to use AI tools to their advantage. The reductionism of this statement is frightening. Not because the statement is fundamentally wrong, but because behind that glimmer of truth lies the fact that the person or company who successfully uses these tools will be able to do the work that currently requires dozens, if not hundreds of people. Employers are not going to pay salaries they can save. They don’t engage in charity, and they are not necessarily bad people for it; they are simply pursuing profit and risk reduction as the capitalist system currently dictates, which we will discuss further later. If you’re reading this and you have the luck of holding a university degree, you might be thinking that none of the activities I just mentioned apply to you and that your job future is secure. You might be a bit more prepared for the future of work, but I wouldn’t rest on those laurels.
Think about how good you are at drawing. I honestly lack talent for drawing, yet there are professionals today who help police investigations by creating forensic sketches of suspects. As difficult as this task is, today these professionals are beginning to compete with technology created by Eagle AI[50], founded by Artur Fortunato and Philippe Reynaud, which offers a free program that, through AI and a very simple graphical interface, allows us to describe a person’s features and in seconds obtain not a drawing, but a high-quality image resembling a photograph of the suspect. In fact, OpenAI, the company behind ChatGPT, published a report in 2023[51], along with the University of Pennsylvania and OpenResearch, estimating that 80% of U.S. employees could be impacted by this technology, with the most exposed jobs surprisingly being those related to mathematicians, accountants, writers, web designers, and lawyers. However, in this particular study, the term “impact” does not imply dismissal, but rather functions that these people perform today and that can already be, or soon will be, carried out by ChatGPT.
White-collar workers with years of professional training should also pay attention to the extent to which technological advancement helps them in their work or actually starts doing their jobs better and at a lower cost for their clients and employers. To illustrate this, it’s worth mentioning recent advancements that have allowed AI programs fed with thousands of cancer patient data to detect some types of this disease more accurately and quickly than well-trained human professionals with years of experience. A trained professional makes deductions based on years of experience analyzing images of various tumors detected at different stages, ultimately relying on their memory, which tends to deteriorate, and their instinct. A software program designed for this purpose can compare our patients’ images with data from not hundreds, not thousands, but millions of patient cases worldwide, in a very short time. This is already happening today, and an example is the use of Watson, IBM’s AI, by the Memorial Sloan-Kettering Cancer Center, which can analyze and compare our data with the medical history of over 1.5 million patients and millions of academic articles from the most prestigious scientific journals[52]. In this way, Watson can not only compare our symptoms but also our genetics with those of other patients, with successful and unsuccessful cases in similar situations, to offer accurate statistics on the best course of action for each particular case.
What to say then about accountants in a world where cash is tending to disappear and all our transactions become electronic, making them easier to audit and track, complicating tax evasion, corruption, and the informal economy. Receiving bribes or setting up a chain of illicit businesses will be more difficult, except of course if one knows how to navigate the world of digital crypto-assets, where privacy and pseudo-anonymity are the order of the day due to the scant and poor regulations of the main actors in the ecosystem. While it’s clear that assets like Bitcoin are impossible to regulate per se by a Nation-State due to the distributed nature of the system, the exchanges of this and other crypto-assets are where the focus should be. Not excessively, not killing this innovative and disruptive industry, but enough to prevent illegal transactions facilitating money laundering, capital flight, and international terrorism financing among other things. Lawyers themselves will have it tough, and it may not initially be due to AI, but due to the automation and digitization of their documents. Today, services like LegalZoom allow people to fill out simple online forms, pay the required fee depending on the procedure, and finally obtain the legal document instantly. People can register their businesses, foundations, contracts, dissolutions, trademarks, make a will, divorce, change their name, and much more, from the comfort of their homes and at highly competitive prices. Not to mention translators facing the advance of technologies that enable real-time translation, not only of texts but also through Skype or Meet, in live audio conversations. Although this technology is not yet perfect, it serves as a window into the future to glimpse what’s coming. I have had the luck to travel to all continents, and while I can usually communicate easily with my Spanish and English, I’ve visited countries where neither of these languages is the main one. Has that been a problem? Never, because the first thing I do upon arriving in another country is buy a SIM card with internet data for my mobile phone. This allows me to access a real-time translator like Google’s wherever I am. This is useful not only for writing and translating words but also for recording our voice or that of another person in any language and then translating it, both in written or audio format, to the desired language. This application even allows us to use our phone’s camera and point it at a menu in a restaurant, a street sign, anything with words in another language, so we can read it easily and instantly.
Regarding the displacement of people from their jobs, studies by the Boston Consulting Group[53] predict that investment in industrial robots will increase by 10% annually in major developed economies, whereas currently, this figure does not usually exceed 3%. In this regard, it is estimated that currently only 10% of jobs that can be performed by robots have already been automated, and before the pandemic, it was estimated that this figure could rise to 23% by 2025.
It’s time we understand that a world with cutting-edge AI is a world where machines and algorithms can perform the four types of work mentioned at the beginning of the chapter. This forces us to reconsider various clauses of our social contract.
What will we do then with this new mass of unemployed people worldwide? Are our own jobs at risk? For now, we should start retraining our workforce, updating education plans to include both robotics and programming, as well as soft or social skills, which may be the hardest to replace due to the logical understanding of emotions and empathy towards others, something machines still struggle to achieve with ease and precision. But beware, never say never. There is a possibility that we might even be forced to change our social and productive organization model, as automation will harm the social fabric if it does not adapt to these changes, given that halting technological advancement is not an option.
With the following statement, I may be jumping several decades ahead, but will we reach the point where work is truly for machines and life is for the enjoyment of humans? How would subsistence look in that societal model? In countries like Argentina, Finland, Iceland, the Netherlands, and Switzerland, among others, there are already discussions about creating a Universal Basic Income as a strategy to defend against technological advancements in the labor market. Even the Economic Report for the President of the United States, prepared by the White House in 2016, indicates that U.S. workers earning less than $20 per hour in 2010 had an 83% chance of being replaced by a machine in the future[54], which aligns with other reports suggesting that nearly 45% of the U.S. workforce could be at risk due to technological advancements in the next 20 years, a figure echoed in academia and one I’ve heard both at the World Business Dialogue in 2015 in Germany and in February 2018 when I attended a conference on International Relations, governance, and technology at Harvard University. Where does all this commotion come from? The starting point of this discussion was given by Carl Benedikt Frey and Michael A. Osborne, two researchers from the Oxford Martin School, who in 2013 estimated that 47% of U.S. jobs could disappear in the next 15 to 20 years due to automation. This estimate is not a definitive statement. The existence of a technology that can replace a particular job does not mean it will happen on an exact date. There will always be political regulations that slow down this transition, though later we will discuss whether that’s good or bad.
At the event held at the prestigious Harvard University, I had the opportunity to talk with Rik Geiersbach, Vice President of Corporate Strategy in defense, space, and security at Boeing, one of the world’s largest aircraft manufacturers. He ventured to say that their most modern aircraft only require the manual work of pilots for takeoff and landing, while the rest of the flight is usually conducted by the central computer under normal conditions. However, he also mentioned that in no more than 10 years, the company could bring to market a technology that could even do without pilots for these tasks. But they have not yet convinced any insurer to provide protection for pilotless smart aircraft transporting people. This is a matter of negotiation, time, and money. Raw, but real.
Reflecting once again on psychologists Wagenaar and Sagaria, regarding the exponential speed of changes and our predictions about them, we have one more example to mention. In 1985, AT&T hired McKinsey to predict the adoption rate of cell phones. The experts from this famous consulting firm concluded that by the year 2000, the industry would reach about 900,000 users. Were they wrong? Yes. By a lot? Yes, by a lot. The number of active mobile phone lines in the United States in 2000 exceeded 109 million[55]. The study was not conducted by a group of primary school students but by one of the world’s most prestigious consulting firms, leading us to ask the following question: What if the predictions about automation and unemployment fall short too? It’s not something I can guarantee, but it’s not something I can refute either. After all, we have established that regardless of the speed of changes, we are not going to stop technological progress.
No nation is ready for the changes ahead. Historically, high unemployment rates lead to social instability, and the lack of consumers in consumer economies also leads to economic instability, a cocktail that is not very advisable. Therefore, we must ask ourselves, what is the purpose of the technologies we are creating in the first place? Technological advancement has its undeniable benefits in education, food, health, safety, and productivity, but perhaps no longer in the current conditions of social organization in the face of this exponential advancement of computers.
It is advisable that, without falling into panic, we begin to investigate these facts and their potential risks, with the purpose of discovering alternatives immediately so that together we can find answers and solutions to these issues. I repeat, the idea is not to worry, but to take action.
At the moment we admit that information processing is the source of intelligence, so some appropriate computational system could be its engine, and we admit that we will continuously improve these systems; while we accept that the horizon of cognition likely far exceeds what we currently know, then we have to admit that we are in the process of building some kind of God. Now would be a good time to make sure that it is a benevolent God we can coexist with.
We cannot close this section without quoting the following phrase from Aristotle in his book Politics (in Greek Politiká), whose relevance does not cease even in modern times:
If every tool, when ordered, or even of its own accord, could do the work that befits it, just as the creations of Daedalus moved of themselves, or the tripods of Hephaestus went of their own accord to their sacred work, if thus shuttles wove and plectrums played the lyre, master craftsmen would have no need of workers, nor masters of slaves.
Thousands of years later, we are still discussing the automation of work, but unlike then, today we have technology that makes it a reality.
[47] Baer, Drake. (2016). “This “Virtual Employee” Is Proof That the Robot Takeover Is upon Us.” Business Insider. Retrieved June 25, 2021, from https://www.businessinsider.com/ipsoft-amelia-profile-2016-4.
[48] Cooban, A. (2023). This CEO replaced 90% of support staff with an AI chatbot. CNN. Retrieved July 19, 2023, from https://edition.cnn.com/2023/07/12/business/dukaan-ceo-layoffs-ai-chatbot/index.html.
[49] Cooban, A. (2023). Shah, S. [@suumitshah]. (2023). We had to layoff 90% of our support team because of this AI chatbot. Tough? Yes. Necessary? Absolutely. The results? Time to first response went from 1m 44s to INSTANT! Resolution time went from 2h 13m to 3m 12s Customer support costs reduced by ~85%. Twitter. Retrieved July 15, 2023, from https://twitter.com/suumitshah/status/1678460567000850450.
[50] EagleAI. (2023). Lab Lab. Retrieved February 1, 2023, from https://lablab.ai/event/openai-whisper-gpt3-codex-dalle2-hackathon/eagleai.
[51] Eloundou, T., Manning, S., Mishkin, P., & Rock, D. (2023). GPTs are GPTs: An Early Look at the Labor Market Impact Potential of Large Language Models. Retrieved March 22, 2023, from https://arxiv.org/pdf/2303.10130.pdf.
[52] IBM Watson Hard At Work: New Breakthroughs Transform Quality Care for Patients. (2013). Memorial Sloan Kettering Cancer Center. Retrieved November 1, 2021, from https://www.mskcc.org/news-releases/ibm-watson-hard-work-new-breakthroughs-transform-quality-care-patients.
[53] Zinser, M., Rose, J., & Sirkin, H. (2015). The Robotics Revolution: The Next Great Leap in Manufacturing. BCG Global. Retrieved November 8, 2021, from https://www.bcg.com/publications/2015/lean-manufacturing-innovation-robotics-revolution-next-great-leap-manufacturing.
[54] Economic report of the president. (2016). [Ebook] (p. 239). Retrieved November 10, 2021, from https://www.whitehouse.gov/wp-content/uploads/2021/07/2016-ERP.pdf.
[55] Andrew Ross S. (2013). McKinsey & Co. Isn’t All Roses in a New Book. DealBook. The New York Times. Retrieved March 8, 2021, from https://archive.nytimes.com/dealbook.nytimes.com/2013/09/02/in-a-new-book-mckinsey-co-isnt-all-roses.