Beware of tech principles in sheep’s clothing

In the last few years, a growing movement has advocated “responsible” technology development and use, particularly for artificial intelligence.

With monikers like “human-centered AI” and “responsible technology” these sound, as we say in America, like “motherhood and apple pie.” Who can be against human-centered and responsible technology?

Companies, trade associations, professional associations, civil society groups, and others have issued 100’s of these statements of principles. The World Economic Forum advocates for “Responsible Use of Technology.”

The Partnership on AI seeks solutions so that “AI advances positive outcomes for people and society.” The Institute of Electrical and Electronics Engineers (IEEE) seeks “Ethically Aligned Design” for AI. These are clearly in contrast to the hordes of researchers and companies seeking unethically designed AI to be used irresponsibly to advance negative outcomes.

At one level, these principles are useful. Of course, engineers and companies developing and using technologies should do so in responsible ways, including ensuring safe products with limited bias and adequate privacy.

But it’s not really clear why developing AI-related products is any different than developing vacuum cleaners and cars. We relied mostly on market forces for companies to produce products that were responsible, because if they did not, they would lose market share.

The problem is that many of these efforts go beyond statements of generally agreed-upon principles to attempts to impose specific, elitist values advancing their particular agenda on technology development.

We see this in a number of areas. Some of these organizations argue that is unethical for scientists and engineers to work on AI that can be used in weapons systems. Who gives them the right to weaken a nation’s defense forces? If a particular engineer does not want to support his or her defense forces, they can work for a company that is not engaged in defense. But it is certainly not the place for the average professional association to advocate for not working on AI weapons systems.

Many of these organizations seek to impose their particular views about privacy on technology. The Association for the Advancement of Artificial Intelligence states the obvious: “AI professionals should only use personal information in accordance with the applicable laws and regulations.” But it goes on to argue that:

Professionals should establish transparent policies and procedures that allow individuals to understand what data is being collected and how it is being used, where it is transmitted for processing, to give informed consent for automatic data collection, and to review, obtain, correct inaccuracies in, and delete their personal data.

But these practices involve tradeoffs that will limit data innovation. If democratically-elected governments want such privacy rules, that is their right. But professional organizations should not dictate those choices.

Where these organizations go furthest in attempting to impose their own values on society is when they pontificate on AI and production efficiency, with many taking a clear stand to not support automation technology.

Emblematic is Stanford’s Human-Centered Artificial Intelligence center which states that: AI should effectively communicate and collaborate with people to augment their capabilities and make their lives better and more enjoyable. Humans are not simply “in-theloop.” Humans are in charge; AI is “in-the-loop.”

They go on to note that: “Scholars envision a future where people and machines are collaborators, not competitors.” One staff person at the Partnership for AI, a group with little or no expertise in economics, writes, when talking about AI that would automate a job:

— One of the things that has become increasingly clear to me is that a lot of the ways that artificial intelligence is being developed and deployed at present are decoupled from what would generate broadly shared prosperity.

— We don’t get much of what we’d be able to identify as genuine productivity increases with benefits that could be spread across the rest of humanity. You probably get some pretty handsome profits for the companies who came up with the products, but you also get very clear, real, human harms on the other side of that.

These harms come in the form of fewer jobs, lower pay for work that requires less specialization than before, increases in physical and mental stress on the job, and decreases in autonomy, privacy, and dignity.

In its report “Ethically Aligned Design” the IEEE states if technology is developed by a few companies in rich countries: The benefits would largely accrue to the highly educated and wealthier segment of the population, while displacing the less educated workforce, both by automation and by the absence of educational or retraining systems capable of imparting skills and knowledge needed to work productively alongside A/IS.”

It goes on to favorably cite a World Economic Forum report that calls on engineers and companies to focus on augmentation technologies, not automation technologies:

The findings of this report suggest the need for a comprehensive “augmentation strategy,” an approach where businesses look to utilize the automation of some job tasks to complement and enhance their human workforces’ comparative strengths and ultimately to enable and empower employees to extend to their full potential.

Rather than narrowly focusing on automation-based labor cost savings, an augmentation strategy takes into account the broader horizon of value-creating activities that can be accomplished by human workers, often in complement to technology …

Not only has this notion that only augmentation technology and not automation technology grows the economy been shown time and again by economic studies to be wrong, these organizations should not try to force their views on others.

If an organization finds that augmenting AI technologies are better than automating technologies, it can make that choice. But if it finds that automating technologies are better, the choice should be theirs, particularly because that choice will boost living standards.

If democratic governments decide they prefer a bit less worker displacement and the consequent slower economic growth that comes with that, that is their choice. It is certainly not the right of well-paid scientists and engineers to impose their preferences on society through professional ethics principles.

In Korea, AI engineers and companies should not adopt principles that constitute social policy operating under the guise of neutral, objective technology principles. Doing so will limit AI development and use in Korea, and slow economic growth and consumer welfare.

Dr. Robert D. Atkinson (@RobAtkinsonITIF) is the president of the Information Technology and Innovation Foundation (ITIF), an independent, nonpartisan research and educational institute focusing on the intersection of technological innovation and public policy. The views expressed in the above article are those of the author’s and do not reflect the editorial direction of The Korea Times.