Is AI/Robotics the Next Humanitarian Crisis?

I am not afraid of technology.  To make technology is to be human. Technology and humanity are inseparable. My concern rests on the agility of our leading global institutions to engage technology in a manner that does not understand other forms of human organization. Technology is human, but it is not innately social. 

Nearly everyday an apocalyptic statement graces the headlines, announcing the threat of robotics to the global economy and patterns of day-to-day living. Supposedly by 2025, one of every three people will lose his job and become replaced by a robot. Five years later, in 2029, robots will have intelligence equal to humans. Artificial intelligence is the “biggest existential threat” to humanity.

These pronouncements do more to demonstrate a poor comprehension of the technologies and a complete disregard for economic history. Thousands of years of evidence highlight technological change is a healthy thing for labor markets. The counter argument is that emerging developments in robotics concerning automation, computer vision, data analysis, and machine learning are giving rise to a new kind of technology that is different from previous developments concerning hardware or information processing. 

The debate on the integration of computation and human society is as old as computation itself, and contemporary arguments remain rooted in the 20th-century invention of cybernetics, an experimental epistemology concerning the effective organization and communication of integrated social and environmental systems. Norbert Wiener’s theories on systems feed back and interactions with technology were situated within concerns of human ecology and governance. Weiner also projected the eventual creation of the machine a’ governor, a robot government designed to solve any problems to emerge within the government-human economy. In consequence, humans would be free to spend time at leisure while the machines replicated themselves, taught themselves, and effectively solved all possible problems for humans. All hail the machine.

According to roboticist Illah Nourbakhsh, in his book Robot Ethics, the critical difference is that while the developments of the 20th century provided the ability to simultaneously engage and manage multiple streams of information, but the advancement of robotics provides the capability to now distribute multiple forms of action. More significantly, the ability for robots to network, sense, choose, and automate with other robots will eventually lead to strange unpredictable configurations of technology that will blur currently held conceptions of social identity and accountability. To render this probability in the context of advanced capitalism, established modes of socio-economic production are at risk of obsolesce. It will be necessary to invent new modes of value production in society or human beings will be useless.

Wiener’s visions were not unique but predicated John Maynard Keynes' Economic Possibilities for our Grandchildren. Economist John Maynard Keynes predicted a technologically advanced future relegating the necessity of human labor to a mere 15 hours a week. There was some plausibility to Keynes' vision, as more efficiency in labor does not replace the need for human labor, but rather elevates the value of human labor relative to the demand for production. Of course, if there is a change in demand, and production is not sufficiently responsive to that change, then human labor is devalued. In the case of robotics, it is imaginable that vast networks of robots producing goods in relation to the algorithmic demands of big data would be more responsive to market changes than humans. Already this is evidenced by the application of algorithmic trading in stock markets. Humans can’t compete with networked, data driven, robots on mere market efficiency. The need for human supplied labor is replaced. The externalities are severe. 

Lost human jobs equates to a reduced tax base, reduced social services, lower quality education, reduced social capital, deteriorating communities, crumbling infrastructure, and stunted generational access to social and personal mobility alongside increased depression and substance abuse. To let this possible future, arise anywhere in the world is unconscionable. Surely this can’t happen of course because humans are capable and responsive. New markets can be created. 

Keyne’s argued that the remedy to competition with technology relies on the ability for humans to acquire the sufficient skills to make use of the technology. Keyne’s highlighted this problem as a “temporary maladjustment:, which might require several years for a labor force to catch up to the technology”. But what if the pace of technological advancement is faster than the ability for humans to learn? To close the gap is an ineffective a pointless conquest. In some places – existing urban concentrations of wealth and opportunity like New York City, San Francisco, Tokyo, or Istanbul – this will be true and educated people who inherited rich social assets will make use of the technology. But no one else will ever have a chance to compete.

Clearly, there is a bigger concern that does not emerge in these conversations on robotic futures: Is a world in which day-to-day livelihoods remain under constant threat of technology a desirable world? As the systems of production, exchange, and valuation that drive the technology are the choices and actions of humans, why then could the long term future of the world be one outside of our own choosing? Obviously, systems magnify behaviors, and complex adaptive social systems are virulent landscapes to contain. The complementary mechanisms and conditions of the global economy, such as systems of governance and human security, have struggled to keep pace with socio-economic demand and it is clear that this trend will only get worse.

There have been tremendous strides in the global economy in the last 100 years. International institutions have been founded to advance regulatory measures for human interests. Less people have died of illness or war. More people in the world have accessed education and social mobility than any century in human history. I am confident in those strides and those institutions. My concern rests on the agility of those institutions to engage a technological threat. The problem of Keyne’s “temporary maladjustment” continues to repeat itself ad infinitum, and when the world places blind faith in the advancement of technology, the learning curve becomes painfully steep.

The international humanitarian and human rights regimes are not contemplating this future impact of robotics. Grass roots organizations and urban planning departments are busy replicating the status quo. In the meanwhile, cybernetics hasn’t died in the minds of military engineers and industrial capitalists remain quick to take advantage of grand visions of a perfectly automated society. We are sold conveniences but in exchange purchase longterm economic servitude.  There is a demand for another approach, other ways for thinking and acting are essential to guide the advancement of human living. Buried beneath the layers of ideology and intent, our globally distributed modes of production and exchange all contain a shared thread and this thread elicits a new possibility for design.

Unlocking Machine Learning with Human Centered Design

Human Centered Design (HCD) is at its core, a process for eliciting the practice of human values and embedding those new values into a new artifact, process, or service.  It is perhaps challenging to consider how HCD may have a role within something as quantitative as Machine Learning, yet HCD can be an important component to the formation downstream benefit for machine learning services. The value to injecting HCD into the design and use of machine learning solutions is valuable to contribute to unknown opportunities and risk the threat of future-tense machine learning decisions.

Quick Intro to Machine Learning
Machine learning is best described as a methodology for a computer program to learn to write new programs without human assistance.  There are two primary domains of machine learning - supervised and unsupervised. Unsupervised machine learning utilizes various statistical functions to identify patterns in information and to extrapolate decisions from that information.  The reference to Big Data is often a use of unsupervised algorithms that are, at their core, a regression analysis. Supervised Machine Learning also leverages statistics, but to do so, relies upon a body of data curated by an individual.  The computer program builds an understanding of the data to replicate it or to use that understanding as a filter on other data.  For example, if 2000 images of a tree are shared with the program, it will use those 2000 images to build a concept of a tree, and will then be able to identify a tree from a suite of images which may or may not contain trees.

Of Machine Models and Human Cognition
To build an HCD approach to Machine Learning, it is important to first distinguish how machine learning is similar and different from Human Learning.  First, both the human brain and the ML program do have a core similarity - they are massive engines of statistical pattern recognition.  Our brains engage and understand the world through pattern recognition. To see an apple, know it is an apple, and understand what an apple does is in fact a massively complex undertaking.  The connection between sight and concept requires our brain to internalize and make sense of about 8,000 different points of information... how light bounces off the surface into our eyes to inform geometry, color, shape, texture, and subtle visual implications of weight, density, and so on that are difficult to articulate.  When our brain has sufficiently identified enough information bits to create a pattern - the apple pattern - it has built a conceptual model.  Our brains are massive repositories of conceptual models, and we use these models to discover new ones. "That is not an apple but it looks like an apple."

Machine learning programs are similar (as most are based on our brain's neural architecture). They use many different methods to analyze information and identify a pattern, like "tree."  Yet their methods are various different from ours, such as rapid quantitative measurements between corners, measurements of curvature, measurement of gaps - and thus generate a very different kind of model.  The ML model is not interpretable to the human - nor does it need to be. The result is the total information required by a computer to make inferences about other possible models.  A computer, also, has limits and it is difficult to make inferences. Thus a human can rapidly intuit a new situation "apple by a tree suggests apple tree," the computer will only see an apple and see a tree,  unless enough information is presented about apple trees to make the leap.

The Learning and Reproduction of Human Values
Machine Learning asserts a transactional value by  using found pattern within a set of information to predict how that pattern is extended against a new inflow of information.  We can use ML to facilitate market analysis, reduce risk in complex situations, and optimize organizations. Yet what does this ability inform... optimize how? Reduce risk to do what?  HCD is a process that requires the designer also conducts pattern analysis - but a key distinction is in the domain of qualitative patterns.  How does a user feel about a given situation over time?  How can that situation shift over time to elicit other feelings?  The intersection of opportunities between HCD and Machine Learning are vast when you consider the bigger context of the data.

Within supervised machine learning, the assembly of training data could be done in bulk or it can be curated.  A designer can pay close attention to the data... what kind of trees are presented? What health are they in?  Can multiple sets of training data be presented that contain healthy trees, sick trees, and variations in environmental conditions?  By curating the data, the human agent can tune the specific types of patterns to be determined, and thus generate greater value later.  Not all data is equal.

Likewise, as the outputs are generated from an ML program - such as new sets of data (often relayed in a dashboard or a customized suggestion), the human experience of that new data point can be catalogued and leveraged to tune the ML program.  While there is such a thing of "Human in the Loop," to aid the precision of the algorithm, another way to thin about this is "Value in the Loop," to reward the program for generating particular kinds of experiences. While abstract, the program does not need to understand the human value - it will build its own conceptual model to make sense of the behavior.  These values can tune the algorithm to new directions over time, wherein we can expect the algorithm to generate new data points very much unlike the original yet of great use.

Geographic Variance in Modelling
A critical aspect of being human is the variations that exists between humans for perception in the world.  Within small groups, language can take on a range of meanings, while across geographies, the variance of meaning ascribed to concrete things can be quite large. Machine learning generates one kind of solution for one kind of person and then generate a new solution for another kind of person... it can be granular, like Netflix, to suggest customized outputs.  Yet what about machine learning applications for broad data sets that are irregular?

Within the design community, I advocate close consideration of the role of place and data. There is a tendency for developers to look to large open data sets. Vast databases exist with training data (like this popular one at MIT).  Yet consistent with my above comments, what are the values which drive the labelling of this data?  For example, if you apply this training data to images of Mogadishu, it will generate results like "earthquake."  Not only is this inaccurate - but the correct label will vary depending on who and where the labeler resides.  A Somali in Nairobi will give a different answer than a design student in Boston to describe a set of data.  Yet most Somalis in Nairobi will supply more similar labels.

Consequently, it is not enough to parse data by meaning.  One must also ground the meaning, and to do this, I suggest the use of classic lat/long coordinates.  If we can build rich databases of place-based data and place-based identification, we can do more than build intelligent softwares, we can build softwares that are flexible to the global shifts in meaning and identity which are traditionally at odds with the demands of computation.  To fuse machine learning with the vast ocean of human value creation and reproduction is a great opportunity for the future.