A firm might satisfy the call for universal sharing of the benefits of superintelligence by adopting a “windfall clause” to the effect that all profits up to some [high amount] would be distributed in the ordinary way to the firm’s shareholders […], and that only profits in excess of the threshold would be distributed to all of humanity.
Nick Bostrom, Superintelligence
Nick Bostrom’s book is full of, very clever, often hard to grasp, ideas around artifical intelligence (AI). What I really enjoyed around this concept of “windfall clause” is that it is actually fairly easy to understand and it is an horizontal concept that can be applied to many things.
Cosmic Endowment
Let’s look, again, into one of Nick Bostrom’s ideas.
In a non-apocalyptic scenario AI can be seen as a tool massively enhancing productivity to the point of making the word productivity obsolete. Indeed productivity is often perceived as how much a worker can deliver, but a sufficient level of intelligence will make most, and likely all, human workers completely obsolete.
This unprecedented extraordinary explosion of wealth is what is described behind the wording “Cosmic Endowment”.
A well-intentionned society could probably fulfill most of our needs leveraging this endowment. In a not-so-well-intentionned society, this would lead to an extreme power concentration where the persons controlling the AI would rip its benefits ; leaving all the others in a very dire scenario, where their skills and work would virtually be worthless.
The windfall clause solves this.
Fixed Market and Damages to Initial Vision
The simple fact of using the word AI puts the whole problem very far in the future. But is it so?
The idea that workers are completely put out of relevancy by machines only starts to become a contemporary debate topic in some limited industries.
However the idea that some companies vastly swallowed their market is a dire reality. And it damages their value to society because competitive incentives do not work anymore:
- A project starts where founders often have good intentions around a project they think is relevant.
- They get investors onboard. At this point current or projected financial criteria become the onlybenchmark for performance. Employees receive similar incentives to “align them” on the shareholders targets.
- The project gains some traction, the problem is getting solved for users and customers, and figures are good enough for all parties. At this point monetization has probably been figured out.
- After sometime the market gets saturated, the project has become a market leader, no competition can fight. Especially on verticals where network effects or captive audience are in place. At this point, the founders, initial investors and incentivized employees probably all had their big win, but they have been replaced by other individuals with similar goals.
- The only way to increase profit is to degrade your product value: charging more for the same.
Incentives are possibly broken here. The invisible hand shows the world an invisible middle finger.
Non-Cosmic-Endowment Windfall Clauses
To avoid step 5, which represents a win for every stakeholders but society at large including customers and users, one could image a similar windfall clauses.
Specifics
The feature of such clauses would include:
- A lower threshold than the one representing some vision of whatever a cosmic endowment could mean, but still high enough to be highly achievable under strong market dominance.
- Some capping on employee and executive compensation, again on an extremly large number. This because shareholders having no incentive to earn more money would have probably no incentive to control labor costs.
- Disposition making the clause unremovable, even with the majority of shareholders. Maybe flexibility could be brought by having third party representatives who would have a veto right if they think society at large would benefit from adjustments.
- Inflation indexing of the above amounts to be able to adjust to a changing world. Possibly expressing them as multiples of invested amounts to not hinder investment too much.
Possible incentives when people are still working
Because AI is not involved in our scenario, employees will probably still be needed. Their incentive as well as the one of investors might be a mix of:
- Minimize their amount of work while maintaining their income.
- Maximizing the durability of their income stream.
- Maximizing their contribution to humanity as a whole.
But purely maximizing yearly growth would suddenly not be on the roadmap.
Possible less desirable effects
The main drawback would be to scare investors away. Because this would effectively diminish the value of moonshots and also the appeal for a potential buyer.