Is Wall Street the superintelligence that will destroy us?

The term General AI might make you think of Isaak Asimov’s “Robot Laws”. Better think about taxing corporations and big government.

If there is a powerful superintelligent AI that is responsible for farming, and it is determined to to fulfill a single purpose, things can go south. A misguided AI that values a high number of strawberries before all other outcomes could lead to some nasty complications. The AI might use too much water or farmland for the breeding of strawberries, leading to a wasteful overproduction. Imagine ridiculous mountains of rotting berries while cattle is dying due to a lack of water.

This idea of an overly ambitious AI, that is stubbornly stuck on a too narrowly defined goal, makes a good dystopian science fiction story or an addictive web game.

Superintelligent “ general” AI does mean — by definition — that the AI is not limited to a single task.

If a developer would want to hard-wire a goal into such a flexible piece of software it must be set into the very core of the software: the reward function. That is the piece of software that tells the software if it has made any progress toward its goal. The reward function is measuring the success of any undertaking of the AI, telling the AI in which direction to pursue further. The reward function does put a value on every action.

Humans have already build superintelligence in the form of corporations. They are build around the common reward function of human society: money.

The most successful companies make absurd amounts of money. We trust corporations more than governments, because we can understand the motivation of companies. They are built to generate value. The programs of political parties on the other hand is usually a vast list of goals which are more or less contradictory.

The problem is: with the money we move to the companies and away from governments, there also comes a power shift. Corporate lobbyist do have a huge say in law-making. This way ordinary people have less and less influence.

Monopoly and Oligopoly

The corporate world has invaded our smartphones, and by that a lot of our attention span. A big part of our life is effectively governed by people who are not elected, but chosen by corporate owners for their business skills.

This is not a good development.

To be successful on a global scale today means you have to find new and efficient ways of exploitation: Mining and drilling for fossil fuels exploits nature. All over the world people are exploited for cheap labor.

The digital world is even more exploitative:

• Smartphones makes it more and more easy to tap into people’s attention span.
• Online marketplaces try to siphon trade margins.
• Intelligence companies, like Google, trade free services for personal data.
• “Incubators” and investors find ways to grab and monetize other people’s ideas.

It is helpful to be aware of these mechanisms.

The world of sales is very much built on the idea, that you can make a good deal. This usually does not mean both sides of the deal are making a profit. Usually there is a side who can snatch more value out of the deal.

To be fair, there are billions of people more on this planet than there was 150 years ago, with better nutrition, better housing and way better healthcare. There are way more video game battles or sports events than real world conflicts.

A lot of this wealth is built on the processing of fossil fuels, large-scale manufacturing and industrialized farming. Nature has almost become a non-factor. But the world we live in is more fragile and more dependent on natural resources than we want to acknowledge. We already have a sand crisis, shortages of clean water and island-sized plastic waste patches in the ocean.

What to optimize for?

Corporations act quickly, which is good, but their short term goals are doing incredible harm. Corporations and corporate AI aim for instant gratification, but they are usually not able to play the long game. Just because it’s very hard to optimize for long term results — the future will always be uncertain.

But money is not all to aim for. All the wealth in the world won’t help you, if there is no planet to live in, no society to thrive in. There is no Planet B yet, and heading for the stars is mostly science fiction. The Moon and Mars are still and will be for centuries very uncomfortable places to live in.

Money must be part of the reward function of a society, but it’s stupid to optimize solely on short term profit.

All those immense computing power Wall Street uses to simply have a look into next week’s numbers should rather be used to solve the important problems we have on this planet.

It’s time to change from specialized AI that focuses on a simple reward function to a multi-reward, generalized AI.

Let’s choose government before billionaires

Actually we have an apparatus that distributes money in a beneficiary way towards multiple stakeholders. It is called government. Yes — my eyes are rolling, too. It’s slow and analogue, it still run with fax machines and forms and stamps. Yes, that’s painstakingly slow. And nobody enjoys filling out ridiculously overcomplicated tax forms. Not even the guys who invented them to make the process easier to file (that’s why there are forms — it’s true!).

But still, government is the way we chose to trust with public money. Because in the end there is a paper trail to every expense and someone who has to take responsibility.

Let’s give the government the tools to speed up the process. Because in the end a politician has to face more scrutiny than any billionaire businessman. Government is not looking by itself for short term money grabs. Yes, people working in government might be, but the more digitized the government is, the more transparency is possible. In theory.

Government is modelling a dynamic world of possibilities with an inert process of lawmaking and a finite set of laws.