The first computers were teams of people, usually women, who did calculations in parallel along with some error checking. The difference in the results between these teams of people and early electronic computers was speed, and even the output of the software on today’s electronic computers could hypothetically be done by people, just impractically slow.
Natural hunger influences my behavior to buy food, but I’m also influenced by the Cookie Crisp commercial that tells me that Cookie Crisp is not cookies but a breakfast cereal and is part of a healthy breakfast, and part of the money I pay for Cookie Crisp breakfast cereal is used to pay for more commercials, lobbying, and campaign contributions. If you consider the millions of consumers, the politicians, the government officials, military leaders, CEOs, workers, etc. and all of the feedback loops and influences among them, it forms a very complex decision making system.
Analogous to the human computers that worked on the Manhattan project and early NASA projects, the decisions of billions of people come together to form our global system. Like the billions of transistors on a chip, or billions of chips in data centers, we have billions of people making decisions that result in action in our economy, government, and social systems. It’s a global AI that’s calculated by billions of relatively tiny human computers.
People trading vegetables and cloth in the town markets thousands of years ago were not planning to create today’s market economy, but now that we have this international capitalist market system, its proponents must defend it. What I was taught in grade school was that the market economy and the profit incentive drove action towards greater and greater efficiency in providing people with their needs and wants. It does seem to work pretty good, and I remember Robbin Williams in a movie where his character, who had recently moved from the Soviet Union to the United States, is overwhelmed by all the different kinds of coffee in the grocery store.
Our socio-politico-economic system is not a pure market economy. A democratic free market country may trade with a country that has a centrally planned economy ran by a dictator or oligarchy, and free market countries will have laws and regulations from central powers. When we get results we don’t like many say that the problem is not enough government control, and another large group says we need more regulation. A smaller group talks about the type of control and effective regulation, but their arguments tend to be more complex, and therefore ignored. There are no purely centralized planned economies either. Countries considered to be centrally planned still have some officially recognized decentralized decision making, and then these restrictive regimes always have a thriving black market.
This giant interconnected system gives us answers, often ones we would feel uncomfortable answering as an individual. How much money should a company spend on safety equipment and training? How many deaths are acceptable when building a skyscraper? How much effort should be made to prevent workers from being impaled by industrial robots? Who gets food? It seems good to get answers while feeling distant from the outcomes, but how do we know if the combination of economics, budgeting, and government regulation is providing the correct answers? Would regulating lobbyists give us better or worse answers? Would more spending on early childhood education? Some other change?
If I think children should not be exposed to Cookie Crisp advertisements, if I think there should be better safety training and more restrictive regulations around the use of safety equipment on construction sites, how can I influence the system to make that happen, but also, should I be allowed to do so? If three people die during construction of a skyscraper, then maybe that’s the optimal number of deaths, maybe the system knows better, and maybe my voice should be suppressed. Many argue that since I have a number of breakfast cereals and brands of coffee in my local supermarket, then that proves that the system is correct, the needs of the people have been met with the greatest efficiency.
The issue of how to guide an AI away from negative decisions is similar to the question of how we guide large corporations, governments, international organizations, markets, and even individuals from making negative decisions. Also, how do we judge what is a negative decision? Some people argue that if the market came up with a decision, then that is by definition the best decision because it met the needs of the market. This is of course a circular definition. A consumer buying a product in a local store will make a decision weighted towards price vs. perceived value, and exposure to marketing and advertising, with less consideration to environmental impact, labor conditions, social justice, etc. Is such a bias good or bad? How do we determine that?
We certainly should continue to explore and investigate the threats that AI will pose on both the micro and macro level, including extinction. This is necessary, but not sufficient. We also need to move from science to applied science. A plan for AI, just like anything else, needs decisions about action, and in this case decisions about governance. I do not believe our current system is doing a good job at properly regulating the human calculated AI that we currently use. If governance in general is having problems, then won’t also governance of AI?
AI research and ideas about AI have been around for a long time. They are here now and need governance decisions now, and we will have to draw on our present knowledge of governance even as we hope to improve that knowledge and practice in the future. Such is the nature of applied science, we can’t always decide to wait for better knowledge or a better technology.
The good news is most of the issues with AI governance are not new issues, and we have thousands of years of examples of good and bad governance to draw from when creating and implementing plans. Also, conversations about AI governance can inform and improve our knowledge of governance in general. As AI governance moves up the list of priorities we should work to make general governance rise with it.