The case for Human Centered AI
It is uncertain how we will learn and work in future, but most contemporaries will agree that change is all around. Whatever your profession, level of seniority or career stage, Generative AI is likely to be a topic of interest or concern for you. The recent AI moment is often compared with the invention of the printing press by Johannes Gutenberg. I think this is an adequate analogy and goes deep. In his book The Gutenberg Galaxy from 1962 Marshal McLuhan writes “We are today as far into the electric age as the Elizabethans had advanced into the typographical and mechanical age. And we are experiencing the same confusions and in-decisions which they had felt when living simultaneously in two contrasted forms of society and experience”. Only 60 years later we are about to transition into a new form of society, experience and work, some call it the AI age. And again rapid change is causing confusions and in-decisions. There might be quick wins but no easy answers. Drawing on my own experiences with digital technology as a physicist, enterprise software professional, entrepreneur and humanist let me try to explain why I think Human Centered AI is the way to go as it both helps to really understand what’s happening and how to thrive both as a business as well as a professional in the face of change.
In physics you make progress not by looking at what’s changing but what stays constant. In contrast the technology industry is obsessed with change. If you worked in IT for more than three decades like me you will have heard these mantras of our profession: change is the only constant, software industry doesn't honour tradition, disrupt or get disrupted, adapt or die, only the paranoid survive. Throughout my career in IT sector I found all of these principles to be true, except that change is not the only constant. To notice this you just need to look up, turn your attention away from tools, metrics and the media.
First and foremost you will notice that people are all around. Or put differently as observed by Jaron Lanier : “Without people, computers are just space heaters making patterns”. My point is that behind every AI model there is a data scientist, behind every AI business an investor, behind every creative content used for model training a human creator, behind every usage metrics a human user. And all of them stay pretty much constant with regards to their needs, hopes, believes, capabilities and vulnerabilities, the human condition or anthropological constants if you will. When I got out of physics and into software business back in 1994 I really didn't know what to expect beyond learning new programming languages. It didn’t take long and I fell in love with my new profession. Not because of the technology but because of the human touch. I remember telling my friends who were still at University that software business isn’t about technology, it’s all about people. This didn’t change in the thirty years since then.
Now look back at all the technology we’ve got but from above. Like Brian Arthur you might ask yourself what technology actually is. During the peak time of the mobile revolution and as product owner of SAP EMR (a first of it’s kind iPad based mobile solution for physicians and nurses) I read his book “The nature of technology”. In his view, technology is not just tools or gadgets—it’s a combination of other, already existing technologies, shaped by human needs and natural laws. What’s interesting is that the nature of technology (like the human condition and the nature of business) doesn’t change. As proposed by Arvind Narayanan and Sayash Kapoor we should look at AI as a normal technology. It builds on top of all the innovations since the invention of the first computer. There is no magic, just math. However what’s special about Generative AI is that it is a multi-purpose technology, a trait it is inheriting from human language. Using the words of David Deutsch, a philosopher and pioneer in quantum computing: it is another beginning of infinity.
So where does this leave us with the future of work in the face of technological breakthroughs? In order to really understand what’s happening I think it is important to try to not be mesmerized by the change at hand (or in the news) because of what Daniel Kahnemann calls the What You See Is All There Is bias (WYSIATI). In November 2023 the world saw ChatGPT and since then a number of very impressive applications of Generative AI technology. As impressive and promising all of this is, it is not all there is by far. Worse, it prevents us from seeing what AI actually is. In the light of the discussion above I suggest to accept the definition given by Jaron Lanier : AI is an exchange of value between people. Once you take this perspective you start seeing the system of systems and the connectedness of people, business and technology. From here it is easy to accept that there are humanists worrying about impact on society, business leaders focused on staying competitive and technologists pushing the limits. In my experience great leaders as well as great companies combine all of these faculties. For me Henning Kagermann (SAP), Steve Jobs (Apple), Marc Benioff (Salesforce.com) and Satya Nadella (Microsoft) come to mind. I was lucky to work for some of them (SAP, Microsoft), partner (APPLE) or compete against (Salesforce). It strikes me that all of these companies are doing extremely well till today despite all the technology changes since I started working. My hunch is that it is because all of them were guided by a timeless compelling vision shared by employees and customers. When thinking about the future of work I believe Human Centred AI (or HAI in short) can be such a compelling shared vision. As pointed out by Simon Johnson and Daron Acemoglu there is no fixed trajectory neither for technology nor the organization of work. Rather there are choices made by individual businesses as well as society. To inform such choices, technology, business and the humanities need to come together. I learned about HAI first time when reading The worlds I see by Fei-Fei Li. She started as a physicist, then turned to computer vision and ultimately AI with stops at Princeton, Stanford University as well as Google. With ImageNet she enabled the breakthrough of Deep Learning and everything which followed. When her mother got seriously ill she started to spend a lot of time in hospitals, appreciating the humanity of doctors and nurses, and at the same seeing countless opportunities to make their job easier (and patients safer) with the help of AI technology. This then became her new North Star guiding her research work, ultimately leading up to the creation of the Stanford Human Centered AI Institute. At the core it’s about interdisciplinary collaboration, bringing together experts from technology and the humanities all guided by the belief that AI must be developed with human dignity, fairness and social benefits in mind.
While I hope that many will intuitively agree to the principles of HAI I can see an equal number of practitioners asking how it relates to them or their business, and overall how this can be potentially operationalized. In his book Superagency Reid Hoffman writes
“With the advances in machine learning that we began to see in the early 2010s, new frontiers suddenly emerged, unmapped and in some ways quite unfathomable. Now, it's a little bit like we're inhabiting the world before Copernicus again, the world before Magellan, and trying to figure out the best way to proceed”
In a very real way practitioners, decision makers and workers are confronted with opportunity and uncertainty, necessity and ambiguity. There are no textbooks and best practices yet to follow. To help them Hoffman is introducing the idea of a techno-humanist compass allowing to make progress while avoiding solutionism (with AI being the hammer everything becomes a nail) and problemism (thinking about AI only in terms what potentially can go wrong). I love this metaphor as it is emphasizing human agency. It is us navigating and deliberating, not suffering and being driven. Like a compass needs a rather constant magnetic field to work (Earth’s magnetic field is flipping on average only every 0.5 million years) the techno-humanist compass needs constants beyond change to work. So how can it be operationalized? I’m afraid a techno-humanist compass is nothing you can buy, download or copy. It’s rather something you need to develop for yourself and for your organization. In my mind such a compass is not a device or an algorithm, and even not text or knowledge. It’s rather an intuition (or culture) formed by an ever increasing understanding (and appreciation) of technology and the humanities.
To bring this to life let me reflect on my own profession and speculate about it’s future. A few days after ChatGPT launched a product manager in my team showed me how he created a spec document with a simple prompt and quipped that product managers probably won’t be needed any longer. Indeed there are voices saying that AI first companies no longer need product managers. All what is needed are sales people connecting with customers and few experts turning knobs fine tuning AI models. I think this is wrong, not out of nostalgia, but because this is not how things (or humans) work. At the core product management is all about vision, empathy, creativity and innovation. By definition none of these skills can be automated. I used to tell new hires that there is no textbook helping to become a product manager. Only way to get there is learning by doing, practicing, failing and growing. According to Steve Jobs it’s not the customers job to know what they want and “Our job is to figure out what they're going to want before they do”. Put differently there is no data or algorithm to automate innovation. Product managers need to be able to put themselves in the shoes of the customers, need to be married with their problems, and not the solutions (or technology) at hand. Sure, AI can help product managers tremendously to become more effective and efficient. AI Agents helping with market research, requirements analysis, product documentation, testing etc. will change the profession by automating tasks which can be automated. But without the human product manager in the loop any company will get at best mediocre products.
What about science? I’m a physicist by education and by heart. Will we see artificial physicists in future? The answer is no, for the same reasons as described above. In “The beginning of infinity” David Deutsch writes
Without the likes of Albert Einstein, Werner Heisenberg, Erwin Schrödinger, Paul Dirac and Niels Bohr there would have been no paradigm shift and no modern physics. Any AI would have fitted the data to classical physics, like many physicists at the time. Nevertheless AI is a truly wonderful tool helping physicists to push the limits and explore the world even further. It even might become an indispensable research tool for fields like molecular biology. But this isn’t different to astrophysicists who are at a loss without telescopes. Without a human scientist in the loop nobody discovers anything.
However there is yet another way to look at the role of AI. In a recent article in Die ZEIT Lambert Tobias Koch, president of the German University Association, argued that AI is not the evil, but a catalyst for much needed change. The way how higher education often works today (students following tight schedules, chasing points and gaming the system using AI) is far away from the original goals of universities. So what if lecturers and students would meet in AI supported discussion labs, focusing on dialogue instead of monolog? Here AI wouldn’t be a replacement for human interaction, but an enabler. I like this example as it shows that change is necessary to make progress. In operations research there is a method called simulated annealing which helps to find a global minimum of a given goal function. It’s name is derived from a method in material sciences to get rid of defects in crystals. Careful and iterative heating up and cooling down of the material allows the system to find it’s perfect (defect free) configuration. To draw on this analogy one can think of professional practices (like education) which are stuck in a moment. People know it can be done better, but don’t know how. Here AI indeed can be a catalyst and enabler to get going again, create new formats, professions and tasks for workers.
To put it all into a nutshell: HAI is a paradigm shift. Similar to what Copernicus has done with the sun, HAI puts humans at the centre, not out of nostalgia, but reflecting realities. There is no technology without vision, and no vision without humans who share it. Instead of dreading the race to the bottom of automation we are invited to reinvent our professions and shape the future of work to the benefit of businesses, workers and learners. If I’d be asked for advice by young people starting their career in the age of AI I would refer them to Steve Jobs who gave a timeless answer to this question back in 2005.
“Your work is going to fill a large part of your life, and the only way to be truly satisfied is to do what you believe is great work. And the only way to do great work is to love what you do”
Stay Hungry. Stay Foolish. Stay Human.