There is a lot of chatter as to the eventual ownership makeup of the frontier artificial intelligence labs and their economic surplus. One question that came up this weekend is whether equity in the companies should be owned in some portion at the nation state level. I am opposed to this for a host of reasons that I’ll try to get down in whatever garbled form.
I do not own a stake in any of the frontier labs other than owning ETFs that own Magnificent 7 exposure who own portions of the labs. I do invest in compute, nuclear energy and cryptography. I believe AI will change a lot about how we do business, my revealed preferences show I live remotely in Montana, I have a tendency toward emergency planning and Plan B scenarios. As a disclosure of my priors.
There are lots of competing interests in this and the self interests from the labs does no any favors. Especially after months, nay years, of overwhelmingly hyperbole about changing labor dynamics, the potential for mass layoffs due to automation as well as obfuscation and excuses about the reason for layoffs in existing companies. And that’s before we get the singularity which is a religious orientation toward making super intelligence that is Godlike in its framing.
I hate this entire conversation on nationalization and socialism. Part of it is that state actors desiring ownership of private companies reeks of the “you didn’t build that” malapropisms from Barack Obama’s presidency in which he attempted to articulate that America’s enormous wealth is built on generational compacts that no one individual could ever own outright. It triggers socialists and capitalists both.
We all contributed in our own ways to the shared infrastructure, institutions, education, cultural norms and the pluralism embedded in our governance systems that enabled the American Dream.
Unless you are a deep partisan, you understand Obama was trying to articulate that none of us made America alone. But the framing from liberals (and populists of all stripes) automatically make this conversation concerning.
Economics is complicated, central planning has a hell of a body count and your average American can only gesture towards the invisible hand and the benefits of self interested commerce. It’s easy to sell us bad policy from envy and fear.
So I must ask why are we acting like we have suddenly won a national level economic boom with clear winners whose spoils must be distributed by the nation state before we’ve even managed to understand how it will be used, at what level an AI model is a commodity and where the benefits will accrue?
Self interested pluralism with a system of checks and balances at the national federal level coupled with states exercising their own interests has been the bedrock of our national success. Changing this has not gone well for us as a nation nor do we have better examples in other nations.
Sure America has had a few twists and turns. The last time we made an attempt at a New Deal post Great Depression worked only thanks to a global world war industrial mobilization in which we won the war and all our other competitors were decimated on bombed our continents across massive geographical boundaries.
And that boom has been largely spent by the children of the generation that fought this war and their children are looking at a pretty significant bill. So why do labs suddenly want to “compensate” Americans and our collective contributions to the models?
And why are politicians taking this bait when we have so little insight into whether we should funnel cash into them in order to own them in trust for some nebulous future?
I have a few reasons in no particular order that I put on Twitter as to why I am opposed to this format of American state equity being the means through which we compensate the people who theoretically trained these models with our output on the wider open web and its content.
1) We don’t know who the winners will be or where the benefits will diffuse (as in post liquidity the current winners might not be the eventual winners) so compensation for model training when the eventual benefits disperse elsewhere isn’t ideal. Why aren’t taxes at state & federal level aren’t adequate enough here should be answered before we make moves
2) Existing IP law doesn’t account well for culture which is a shared co-creative process (i recommend Susan Scafidi of fashion law institute “who owns culture” ) so compensation is already not easy to track back
3) A state entity w the monopoly on violence can do a lot of damage on the margin by not fully understanding who created what and where it is applied especially in non deterministic systems
4) Much of what the models were trained on was open source licensing including the company where my own family made money Stack Overflow. We got paid sure, but none of this would exist without the effort of its users who contributed on those open license terms. But clearly the final value of the content created & company’s value itself were harvested much further down the line in enterprise contracts for coding models. It was not in the management of an open source license community product created by users or managed by engineers, so who should have been paid? The users who wanted their content to be open sourced? The volunteer moderators? The full time employees? The shareholders of the company, the buyer of the company or the users of that data set at Claude or Cursor or OpenAI? Or is it Americans that never even heard of SO? Where does value accrue over time versus point in time? It’s not an easy question to answer is it?