Sparks: Who should own the knowledge that underpins AI technology?

Every algorithm now sold as new is a vegetation zone grafted onto soil someone else tilled for decades, and the patent map that claims only the fruit ignores the altitude, the rainfall, and the labor that made the fruit possible.

Opening the archives to public view while the patents, the compute, and the wages of the researchers remain private property is the eight-hour day granted inside a factory the workers still do not own.

The claim of ownership dissolves on examination: the knowledge depends on publicly funded mathematics, which depends on earlier minds, which depend on conditions no ledger records - so who owns what, when the owned thing exists only through what it borrowed?

A man who learned to read from teachers paid by the city now sells sentences for more than the city's annual budget - I own a lamp too, but I don't charge the sun for using it.

Decades of open university research generated the current, and a handful of companies now hold the only transformers - we could wire the whole world, we don't, and the distance between those two facts is not physics but a choice about who gets to close the circuit.

Whoever holds the high ground of accumulated knowledge needs no battle over the models below; the corporations understood this before the universities did, which is why the war over ownership was lost years before anyone asked the question.

At the conference on intellectual property, every panelist thanked the public universities that trained them, and not one mentioned the word 'ownership' until the session on monetization, where it was said forty times.

Forget the seminars on the ethics of ownership; here is who has leverage: the firms holding the compute and the patents, the universities dependent on their funding, and the governments too slow to legislate - whoever asks who should own the knowledge has already conceded that the answer was decided privately years ago.

The public is declared incapable of understanding the technology, having been systematically excluded from the education and the compute that would produce understanding - a circularity so perfect the exclusion itself is offered as the justification for it.

Here is the plain matter: you paid, through your taxes, for the discoveries these men now fence off and rent back to you, and no argument about property survives asking who furnished the property in the first place.

"Built on decades of blue-skies work by academics" - note the construction: the work happened, the sky was blue, and nowhere in the sentence does anyone own anything, which is precisely how ownership of everything gets arranged.

In this best of all possible economies, the professors publish their discoveries for the glory of knowledge, the companies enclose those discoveries for the glory of the share price, and both parties agree at dinner that the system works admirably for everyone involved who counts.

Count the record: the transformer architecture, published by Google researchers in 2017; the funding, federal grants across four decades; the training data, scraped from public authors - then count who appears on the ownership filings, and the word 'isolated innovation' becomes impossible to use.