Ethics and Technology: From Neutral Infrastructure to the Age of AI
From its inception, going back to the postwar (World War II) period, the technology industry was heavily involved in the design, development, manufacture, and sale of infrastructure. Initially, the infrastructure was vertically integrated hardware and software. More recently, software has come to predominate, with the rise of disaggregation, where hardware substrates were decoupled from software-driven intelligence.
The technology industry is notoriously inept when it comes to naming its own artefacts. That’s how we found ourselves with “software-defined networking,” a mechanistic construct, describing, at best, how networking technology is put together; it doesn’t tell you anything about how or why it emerged or what purpose it serves. In reality, the software itself didn’t so much redefine and repurpose the network as it facilitated a new form of networking.
Developers, building a new wave of business applications in clouds and in on-premises datacenters, were the ones who forced and drove major changes to infrastructure. They needed networks to be more agile, more flexible, more elastically scalable, unconstrained by lengthy and decidedly non-agile hardware lifecycles. Consequently, “software-defined networking” was a misnomer. The network’s new requirements were defined by developers, and software was the necessary means by which those requirements were satisfied.
The industry should name its creations with more rigor and precision, but it doesn’t. We’re left in a constant cycle of revisionism, retrospectively trying to put right the sins of exigency-propelled imprudence. Things need names, and there’s pressure to come with something fast. The problem is, we’re forced to live with the result, at least until posterity corrects our haste.
Still, a network increasingly composed of software, driven by the needs of developers and their applications, remains infrastructure. It is a form of plumbing, albeit increasingly automated and sophisticated, less refractory and uncooperative than its forebears.
The Indifference of Infrastructure
Neither plumbing nor infrastructure has an abstracted, financialized intent. It is not tendentious; it can be used as a substrate, or foundational layer, by any cloud or any business. It doesn’t care what applications and data ride above it, though it must be sufficiently capacious and flexible to accommodate whatever is specified by changing and varied requirements.
Infrastructure providers, and their employees and investors, can justifiably claim a sort of neutrality. Infrastructure is required by governments, by businesses (legitimate or not), by any manner of organization pursuing digitization. The infrastructure provider merely supplies infrastructure; it is not trying to define the business or service layer of the modern business, economy, or polity. That is up to others, namely those who buy the infrastructure and deploy it as they see fit.
That is as it should be, though the technological Cold War between the U.S. and China has resulted in a growing geopolitical prohibition on what can be sold to whom. These are games above infrastructure vendors’ heads, and there’s nothing they can do about it other than follow the rules.
Infrastructure is not existentially endangered by artificial intelligence, now or in the future, no matter how far into the future you wish to look. Nevertheless, infrastructure will continue to proceed along a path, which it has been on for more than a decade, toward greater automated intelligence. AI will play a functional role in accelerating this trend, investing the network and other infrastructure elements with greater operational autonomy, not only before deployment but afterward, in day-to-day management. Fewer employees will be needed to manage and maintain the infrastructure. We should all be honest about that outcome. It’s already happening. There are fairly large organizations where only a handful of people, if that, manage most infrastructure. These organizations embrace intelligent automation unreservedly, and others are following their leads. The business benefits and value derived from taking that route are compelling, and modern organizations are drawn ineluctably to cost savings and increased efficiencies.
A larger point I’d like to restate is that infrastructure, in and of itself, doesn’t do any harm. As I said, it’s neutral. I suppose you could argue that it does environmental harm, and that its emissions could be reduced to improve sustainability. I accept that assertion. Other than that, though, infrastructure doesn’t have an innate opinion or a bias toward one another or service over another. It can support nearly anything, and the vendors who provide infrastructure probably shouldn’t have a view, within reason and legal parameters, about what exactly customers choose to run over it.
As for those customers, some of them might wish to subject themselves to greater introspection. We have reached a technology epoch in which the largest organizations – increasingly digital, and heavily abstracted and financialized from the old economy of commodities and tactile products – have attained rarefied influence and power over those who use their services, and even over modern governments, mass media, and other major institutions. These abstracted and financialized companies, including the largest of social-media purveyors, are necessarily biased toward their own interests, and they wish to exercise their power unreservedly, often with no accountability or responsibility for any damage, collateral or otherwise, that they cause. Technology companies have never had such power before.
New Age, New Ethics
As human beings, we are more than employees. Most of us lead compartmentalized lives, with a professional persona and a personal persona, one for work and one we present to family and friends when we’re off the clock. We might have other personas, but we all have at least these two.
When I was employed some time ago at a company involved with grid computing, a forerunner to cloud infrastructure, nearly everybody who worked there knew that some of our customers represented businesses and industries of which we might approve ethically, while other customers were involved in industries and practices with which we might choose to avert our eyes and limit our thoughts. That was our prerogative, I rationalized, because we were just providing infrastructure. To the best of our knowledge, those organizations were not doing anything illegal. They were businesses and organizations operating within the law. Besides, we weren’t deciding what to do with infrastructure, how it should be employed and what higher (or lower) purpose it should serve. The customer made those decisions.
Even so, I often thought about the employees of some of those customers. Did they choose to work for those organizations, or did they have no other recourse to gainful employment? How did they reconcile their professional selves with their personal personas? Did they even consider these questions? I would also ask: Was I being unfairly judgmental?
I suppose how we answer those questions, if at all, is a function of personal choice, how each of us views our place in the world. I have a distinct feeling, however, that vexing ethical questions will become unavoidable for many who work for the largest of our AI-powered tech behemoths, seemingly destined to exercise unprecedented influence and power over people, countries, institutions, and industries.