My take on AI Alignment: Corporate misalignment and DAOs

[act65]

In this post https://​​act65.github.io/​​alignment/​​ I go over the relationship between corporate and AI alignment to the ‘public good’. Concluding;

Our ongoing failure to align corporations with the public good is a live demonstration of our inability to solve alignment problems.

And in the follow up post https://​​act65.github.io/​​align-dao/​​ I go over how (I think) democratic processes via DAOs might be the solution. And propose a framework for analysing DAOs alignment.

The only people who can define the public good are the public.

[Karl Krueger]

The modern corporation was born from a powerful idea: to pool resources and limit individual liability to achieve great things for society. Imagine building a transcontinental railway in the 19th century. The project required a colossal amount of capital that no single person could provide and a level of risk no single person could bear. The corporation was the mechanism that made it possible, allowing thousands of people to invest their resources toward a shared, progressive goal. It was a tool for progress, designed to align private investment with the public good.

[...]

But somewhere along the way, the original purpose of the corporation became warped. The singular goal of maximizing shareholder value became the law of the land. This wasn’t an evil plot; it was a shift in incentives.

Those transcontinental railways weren’t the first joint-stock corporations with limited liability. Colonial enterprises such as the British and Dutch East India Companies and the Hudson’s Bay Company were much earlier.

And those were intended to align the private interests of investors with national expansionist interests. The idea was to use private investment to expand the national colonial project beyond what the treasury could afford. These enterprises were granted legal monopolies on trade with their colonies; and soon developed their own military capacity, both to impose their rule by force on their foreign subjects, and to defend themselves from international rivals, pirates, and so on.

Put another way: These early corporations did not aim at “a shared, progressive goal”. Their goals were, from the beginning, both rivalrous (each European nation against all others in the scramble for colonies) and extractive (seeking the good of European investors and rulers at the expense of the colonies’ foreign subjects).

From the standpoint of the CEV of humanity as a whole, these early corporations were no more aligned than modern AI companies, and arguably a lot less so!

[act65]

Ahh, thanks. That changes the narrative a bit doesn’t it.
I’ll need to reframe the argument slightly.

And I wasn’t trying to say that early corporations were more aligned. Rather that it was intended that they were aligned (tho, as you say, this seems wrong!?), and now corporations know how to hack the rules better.

[Trevor Hill-Hand]

I have had this same feeling, in these same words, for many years now.