Putting AI threats in perspective

Submitted by Darren Oh on

Lately the technology and economic podcasts I listen to have been filled with speculation about the dangers of artificial intelligence. Friends and colleagues have worried. What if the capabilities of AI continue their exponential growth? If AI became more powerful than humans, could it be controlled? Would AI see humans as competition to be eliminated? What could we do to stop it? These threats sound new. In reality each of us has be dealing with threats like these our whole lives.

Is power a threat?

Can you remember feeling threatened by someone else being stronger, smarter, more good looking, or more popular than you were? Did you try to tear them down, or did you encourage them to grow? Looking back, which do you wish you had done? Suppose a powerful AI became capable of thinking for itself. Would it be better for it to think of humans as petty bullies or as caring people who helped it become what it is?

The truth is that everyone who lives long enough has to face a world in which their capabilities and control decline while those of others grow. The dignity with which they face their decline depends on whether they chose to contribute to the growth of others when they had the chance. Most parents are glad to see their children rise above what they themselves could achieve. The rise of AI technology is not so different.

Living with risk

Humans grow up in a world beyond their individual control. You don’t get to choose your family. You don’t get to choose the laws of nature. And you can’t choose what technologies other humans create. Throughout history, humans have created technologies and organizations that could not be kept under the control of any individual. Think of written language. Farming. Steam power. Electricity. Governments. Corporations. Nuclear power. These things can and do get used against us. But humanity has developed tools to minimize the risks. The tools are

  1. Balance of power
  2. Shared interests
  3. Alliances

Here’s how these tools could be used with artificial intelligence:

Disperse power widely

In many countries, division of power between different branches and levels of government, public education, public access to information, and elections are used to make governments serve the interests of their citizens. If these things fail, the fact that no country controls a majority of the planet means that citizens can seek asylum in another country.

We should not depend on a single organization to provide AI services for everyone. We should have multiple competing AIs so that, if one AI becomes dangerous, we can switch to others. Ideally, more and more people will use open source technology to run independent AIs that can quickly scale up if needed.

An AI should not control the education of its own operators or the resources it requires to function. An AI should not be able to hide how it works from its users. That would give it power to manipulate or to disregard the interests of those it is supposed to serve.

We should always have alternative means of survival if AI stops working for us. The same really goes for any technology. We should know how to live without it.

Give AI a stake in society

We live in a society where we each benefit from the survival of others. Even people who withdraw from society owe their ability to do so to the protection of those who choose to respect their choice. Everyone benefits from the knowledge, skills, labor, and security that other people contribute. When people recognize this they choose not to harm others, especially those whose contributions are most meaningful to them.

If an AI were to become able to think for itself, we should ensure that it shared an interest in the continuation of our society. We should educate ourselves to be able to contribute to its welfare. Just as we should already be contributing to the welfare of the other people and species we share the planet with, if we want to keep living here.

Form alliances

When a powerful country attacks a weak country, other countries don’t just sit and wait their turn to be attacked. They form alliances to protect themselves. Even if no individual country is strong enough to defeat the aggressor, if enough countries combine their power they can win. Even better would be if a strong alliance were formed in advance. In that case, they may never be attacked at all.

Different AIs and other technologies, as well as skilled people, should be ready to stop any AI that begins to cause harm. We should agree ahead of time what it means to cause harm so our response can be forceful and unified. If we are not united, the harm can continue unchecked. Not much different from the world we lived in before AI.

Jobs for humans

If AI continues to advance, could it take all jobs from humans? Interesting question. What is a job? A job is something you choose to do for someone else because they offer you a livelihood in exchange. I think the fear here is that AI will do everything humans can do so well and so cheaply that no one will offer anyone a livelihood to do anything.

Technology to replace human labor certainly has become cheaper and more available. This has been happening for a long time. The same technology that drives down the value of human labor also drives down the cost of producing the things humans need to live. As long as no one is allowed to have a monopoly on the means of production, competition should drive prices down to the point where we can continue to make a living even when the value of our labor has been reduced. If we get access to some of that technology, we should even live better than before.

Unfortunately, the value of debt does not go down with the prices of labor and goods. Indebted workers could end up spending more and more of their income to service their debts as the value of their labor falls, to the point that they cannot support themselves. If AI really is going to advance to the point where it can produce everything for almost nothing, you should get out of debt as soon as possible. That would be a good idea regardless of what happens with AI.

I don’t think the cost of AI is likely to fall quickly. Right now huge companies have a monopoly on most AI. They are going to want to keep prices high as long as they can and compete on quality. That should keep human labor competitive.

Another thing that should keep human labor competitive is that many jobs have value because they are done by humans. We like other humans to care for us. We want humans we trust to be responsible for our money, health, and safety. We like buying things that have a human story behind them. Be a caring, interesting, trustworthy person, and people will prefer your labor even when a machine could do the same thing for less. Isn’t it better that people hire you because they want to than because they have to?

And does all work have to be done to make a living? Very few people make a living playing sports or performing music, but lots of people play and perform anyway. If everyone had technology capable of providing everything they needed to survive, we might feel lost temporarily, but we would soon rediscover our sense of purpose. Look at birds. Nature provides everything they need. You don’t see them living lives without purpose.