Valuing Humans in the Age of Superintelligence: HumaneRank

Did you receive this forwarded from a friend?

What happens in a society where nearly all human intellectual output is out-priced and outperformed by AI?

We're human, and we need humans to have value - economic and otherwise. A world where humans have no economic value is very dangerous for us. It sets the conditions for some extremely grim outcomes. For example, we should be very concerned about the actions of a national leadership over citizens that offer no economic value.

The problem? I am absolutely convinced that we are headed for such a world.

To believe this claim, you only need to believe three things:

1. Our AI systems speeding towards and past various measures of human ability will get there,

2. That the cost of AI performing most human tasks will be significantly lower than existing humans performing those tasks,

3. That employers and buyers will choose the lower cost options.

For the purposes of this piece, I will assume the claim is true: that nearly all human intellectual output will be out-priced and outperformed by AI, resulting in most humans having zero economic value.

UBI is a stopgap, not the answer

Universal Basic Income (UBI) has been put forth as a solution to AI-driven job displacement, but it only deals with half of the problem created by that displacement.

Let's assume that a UBI policy is implemented, at least for documented cases of AI-driven job displacement. The next equally pressing problem on our hands is finding an activity for AI-displaced people, ideally a productive activity. UBI provides recipients with a minimal claim on goods in the market via money, but UBI can't make the human recipient economically valuable.

As a crisis management tool early in the era of AI job disruption UBI may work, and is probably necessary.

The answer could be leisure time - but the history of idle males especially points to the danger of not giving people something concrete to do with their time.

The timing of a solution to this issue is critical. For that reason, I believe we need a buffer occupation that people can enter by default, at least until we determine a mechanism for allocating human effort to broader areas. What could the buffer occupation be?

I will say at this point that one of my own requirements for a solution to this problem is to preserve human freedom. Humans need not just to have value, but to remain agents with the freedom to decide their own actions in the market and otherwise. And that means, simply, that we need to prevent a slide into a communist dystopia.

In the world I’ve described, the only human output with value becomes intangible outputs. And those intangible outputs can only be valued by other humans.

Ultimately, we need a way to:

• distribute claims on the surplus of society, and

• maintain a price mechanism

The actual proposal is easier to grok if you know a bit about PageRank - it’s the idea that launched Google. Let’s do a refresher on it here - if you’re already familiar with the idea behind PageRank, feel free to skip this section.

First - a short backgrounder on PageRank

The web boom begins

As the web grew exponentially in its early days, the problem of page discovery became critical. Up until that point, people relied on human-curated directories like Yahoo to discover new web pages. But this system had a few critical flaws:

1. a new page could take a long time to appear in the directory while it (a) waited to be found by somebody, (b) who then needs make the effort to submit it, and (c) waited to be reviewed by a directory moderator who chose whether to accept it, and what categories it should appear in

2. moderators were just individuals with opinions - why should their opinion of a page be valued over the opinion of people on the Internet broadly?

3. pages that made it to preferred positions in the directory often enjoyed the benefits of that position even after their content became stale or unmaintained

It became clear that an automated solution was needed in the form of search engines. But the problems of creating the search index efficiently, ranking every page on the web, and returning results quickly represented a set of unsolved computer science problems.

Google solves the search engine problem

There were lots of search engines before Google, so why did Google win? Two reasons: one computer science advance, and one business advance. The business advance was the creation of the AdWords pay-per-click advertising platform that drove Google’s incredible earnings growth. The computer science advance was PageRank. Note that the “Page” in PageRank doesn’t refer to web pages, but rather it’s creator Larry Page, founder of Google with Sergey Brin.

PageRank is the name of an algorithm for determining a fair ranking of web pages based on the links between them, and also used to refer to a web page’s ranking in the system.

Each web link is treated as an endorsement of the destination page by the source page. The amount of “endorsement credit” that flows due to each link depends on the PageRank of the source page, and how many links the source page made and hence divided it’s endorsement between.

Because of how the rank value is defined, it requires solving a very large set of equations to determine the PageRanks of all the pages.

In the end, the great insight and benefit of PageRank is that it doesn’t impose a ranking on pages based on a particular person’s opinion. The rankings are derived directly from the importance of each page as implied by the pages that choose to link to it.

Now that we have the prerequisites, on to the main proposal.

A Proposal: HumaneRank

Imagine a system in which every month, every human in the nation state or other political polity participates in an exercise of the grant of endorsements to other humans. And each human is empowered to grant their endorsements to any other set of humans, in whatever proportions they see fit.

To make the mechanics simple, granting happens through an ultra-basic voting system - a spur of the moment digital bump, via whatever UI you like. To make things concrete, just imagine you can point your phone and tap to endorse a person. Ungranted endorsement flow from a humans has no effect for the period (i.e., the unused “flow” does not accrue to the human that didn't grant it).

At the end of the month, each human’s endorsements are expressed as links to the endorsed humans, weighted by the number of endorsements they gave each human. If I give 8 endorsements to Bobby and 2 to Andy, this will be expressed as a “link” with 80% of my endorsement flow to Bobby, and 2% flow to Andy.

Now we run the PageRank algorithm on the population. We end up with a score for every human in the political polity.

Each human's HumaneRank at the end of the period determines their claim on the resources of the nation - in the form of good old money. We distribute the monetary surplus of the nation in proportion to the scores. Anyone who receives zero grants for the period still gets a Universal Basic Income for the period.

Before concluding what you think of this system, consider the following desirable attributes of this system.

HumaneRank has a number of very desirable qualities:

1. It preserves the price signal mechanism in the money economy, ensuring that market efficiency is maintained throughout the productive (and heavily AI) economy. It does not put fingers on the scale of the market mechanism.
   

2. It creates a new sort of price (happiness) mechanism in the granting of HumaneRank endorsement credit by humans to other humans.
   

3. The points system substitutes for the valuation of each human’s output in the money economy, in which that output would otherwise have zero economic value, our original problem.
   

4. It is naturally resistant to criminal activity:

• If you scam me by tricking me somehow into granting you my points for period P, I'm certainly not granting you any points for future periods P+n

• Further, it's not clear exactly what you could do to get me to grant my points to you. For one thing, if I digitally bump you N times, there is nothing to prevent me from bumping someone else 10xN times right afterwards. Besides, the “duress” scenario is fairly pointless.

• In the Google version of PageRank, the way people typically scam the system is by creating a bunch of spammy websites and point their links ("points") at the site they want to prop up. But in HumaneRank, that would require creating humans - not possible.

5. It is a form of iterated game. Iterated games (in the sense of game theory) are known to promote pro-social behavior.
   

6. It preserves human freedom. Every human is free to direct their endorsement flow to anyone, or any group, in any proportion, for any reason.
   

7. Very interestingly, the motivation to grant your endorsement to someone is basically whether they have provided you utility, or made you happy in some measure.

Stumbling on this solution to the problem of the economic value of humans post-AI has improved my outlook on the issue. I do still believe that left to a solely core (money) economy, all human output will be valued at zero due to being outcompeted by AIs. We need dramatic new policies to avoid grim scenarios, while not sliding into the dystopian politics of the past.

If you liked this issue, I’d appreciate if you would share Road to Artificia with a friend or colleague!

Share by

• forwarding this email

• sharing a link

• or clicking the share button below

Thanks!