Artificial Intelligence and the Future of Work
I propose to consider the question, “Can machines think?” […] “Are there imaginable digital computers which would do well in the imitation game?”
Turing, A. M. (1950). Computing machinery and intelligence. Mind, 59(236), 433-460.
In 2017, thinking machines are at the core of progress and change in the global economy. It comes as no surprise that artificial intelligence (AI) is also a recurring theme in the Marshall Memorial Fellowship, leading from the U.S. East Coast to the West Coast. With advances in machine learning, enabled through large datasets, computing power and algorithmic innovations, technology can help us to create systems that are able to outclass human-level intelligence. While modern computer’s abilities have often been showcased in games such as Chess, Go, or Poker, computer intelligence has infused areas such as medical imaging, speech translation, vision, driving, power efficiency, movie recommendations, and other human-expert capabilities. As robots learn to hear, see, feel, and reason, technology plays an increasingly important part in solving today’s problems.
Machine learning, the core discipline of AI, is also a very good example for the American “Miracle Machine”, as Eric S. Lander and Eric Schmidt describe the ability to spin-off funded research. As the authors write, “[…] When scientific breakthroughs spawn new industries and jobs, those benefits occur right here in the United States — because companies want to remain close to the flow of new discoveries and experienced workers.” In exact these days, we can observe how laboratory curiosities of the 70s and 80s, neural networks and deep learning, pave the way for services with significant commercial value, changing the way labor is deployed in our markets.
Open online education, open-source software, and experiments, provide the next generation of technology entrepreneurs with tools to create novel robot ventures. Noteworthy in time of immigration debates and gender equality, thousands of smart immigrants from around the world and several women in AI research make a significant contribution to the AI ecosystem. While current AI systems can be described as “Vertical AI” (or “Narrow AI”), that means tools focusing on very narrow tasks, as they could be performed by a human, computers are slowly advancing to the next stage. While the thought experiment of the Turing Test will occupy us for some years to come, computers are soon able to process context and knowledge about the world more broadly. It will add up to something like what we describe as “common sense”. Such artificial general intelligence (AGI) describes the intelligence of a machine that could perform any intellectual task that a human being can. More or less, the dichotomy between the two are pre-programmed and human-like self-learning softwares that run such systems. The current development foreshadows future companies, able to coordinate workflows and shop-floor robots in a single line of code; fully automatic organisations. Probably inevitable, such a scenario comes close to a Kubrickian idea of the future and herein embedded angle on technology that puts human workers in jeopardy. Though the true future may be a more human, the biggest public policy (but also private companies and investors) challenge might sound: How can we make the lever of technology work for people, and not against them?
“As leaders, it is incumbent on all of us to make sure we are building a world in which every individual has an opportunity to thrive. Understanding what AI can do and how it fits into your strategy is the beginning, not the end, of that process.”
Andrew Ng, Co-Founder of Coursera; Associate Professor of Computer Science at Stanford University Source
A Challenge to the American Dream
As computers advance in the “Imitation Game”, the concept of inclusive innovation has never been more critical. Noam Chomsky underlines in the documentary “Requiem for a dream”, one core element of the American Dream is class mobility. Indeed, Ray Chetty and his colleagues recently find that the rise in inequality and the decline in absolute mobility are closely linked. As the researchers argue: “If one wants to revive the “American dream” of high rates of absolute mobility, then one must have an interest in growth that is spread more broadly across the income distribution.” The question arises, how AI will influence the distribution of income and herewith social mobility? As economists Erik Brynjolfsson, Andrew McAfee, and Michael Spence frame it, those who can innovate and create new products, services, and business models, will reap huge rewards. As the scalability of computing today enables a very few to deliver services for millions of people, the distribution of income for this creative class might take the form of a power law. A small number of winners capture most of the rewards and a long tail in the society might be left behind.
Analyzing industrial robot use between 1990 and 2007, the economists Daron Acemoglu and Pascual Restrepo recently conclude a negative and significant impact of robots on employment and wages. On the other hand, authors like David Autor, argue that economic history shows that automation not only substitutes for human labor, but complements it. Below the line, it would be foolish to predict which school of thought is correct this early in the game. However, with regard to social inequality, Silicon Valley is very aware of the ripples that new technology can bring throughout society. Recent debates in the Bay Area, among other innovation hotbeds, are driven by discourses and experiments on universal basic income, negative income tax, or a tax for robots. Creators and investors seems increasingly aware that innovations must be inclusive in order to be socially accepted.
Experimentation and innovation are constants in human history. So are fear and insecurity when it comes to new technologies. In my point of view, the hypothetical, long-term concerns are outweighed by the possibilities. That being said, hearing the critical voices is necessary. Understanding and addressing the societal challenges brought on by rapid technological progress remain tasks that no machine can do for us. While technology will be a part of many current and future solutions, technology alone is not the solution. Will artificial intelligence have adverse effects on the employment landscape? Of course it will. Do thinking machines have the potential to improve infrastructure, mobility, and health services? You bet. Will it create new jobs and ecosystems? Here is the silver lining: Certainly.
Change is always bittersweet. However, we should not think about racing against machines, but with them. The important question is not when machines will surpass our intelligence, despite the question how we define “intelligence”, but how we can work together with them in new ways. As progress will happen incrementally, and our task is to make sure that all parts of society can keep the pace. Ultimately, technology does not want anything. As its rationality is bounded, AI depends upon the human interpretability of algorithmic decisions, but also sentient and creativity. As latter is hard to encapsulate in code, we have to decide when and where humans need to be involved in decision-making. Colleagues at the MIT Media Lab call that “society-in-the-loop” artificial intelligence. As such, technological innovation has to be accompanied with societal innovation and both vertical and horizontal mobility in our educational system.
Looking at our human history, technology has always been an enabler. As AI pioneer Marvin Minsky notes in 1994:
“Whatever the unknown future may bring, already we’re changing the rules that made us. Although most of us will be fearful of change, others will surely want to escape from our present limitations.””
Minsky, M. L. (1994). Will robots inherit the Earth?. Scientific American.Read More →