Divisionmaxxing

Divisionmaxxing

As our economy becomes more and more driven by intangible assets, the intelligence industry divides our society in a new way.

There is no way to measure or even define intelligence; no one dimensional way to score a human being’s capacity for productivity. Systems or tests attempting to measure something called intelligence are fraudulent and frequently racist also.

There are entire books and papers written about why measuring intelligence doesn’t work, but the recently inability to even measure much simpler AI agents gives us the word benchmaxxing:

the practice of excessively optimizing artificial intelligence models to achieve high scores on public leader boards and standardized tests rather than improving real-world utility

Benchmaxxing goes layers deep because AI agents are optimized to take coding tests and coding itself is usually failing to improve real world utility.

The rules of our current economy were written mostly by the needs of colonialism, slavery, and manufacturing — nation states invest internally in elites and, where possible, steal from abroad. Now these same rules are being applied to an intelligence industry.

As bad as colonialism, slavery, and unregulated manufacturing were they all had in common that there would be winners. Colonialism moved and still moves resources from a global south to a global north. Slavery helped build America into super power (and even built the protective wooden wall that gave Wall Street its name). Unregulated manufacturing, as bad as it was for the environment and workers, does ultimately produce things although occasionally those things have ambiguous utility like tobacco.

An intelligence industry has more ambiguous utility all the time. Its output can be propaganda or spying or inventions that cause net harm.

People who go to work with people of a different race or gender can see for themselves that the supposed productivity gains between them are a sham. You can form a racially or gender pure group but it will handicap your endeavor. You can try to force everyone to one religion, but eventually that religion will splinter.

Intelligence is different because it sweeps so many variables under its rug. In one form intelligence might be just adults raised as privileged children. In another form intelligence might be a gift for mathematics or eloquence. Gather everyone together that shows a spark for anything, call it intelligence, and then build your economy around rewarding these people while shaming and devaluing everyone else.

Because intelligence is so undefined Elon Musk can be the child of privilege and also pretend to be an engineer. He can be an intellectual but also an anti-intellectual. He is by halo effect the smartest at everything but also clearly insane.

This complete lack of definition was previously the stuff of gods. Jesus is descended from king David and Abraham through Joseph, but also of virgin birth and not blood related to Joseph at all. Caesar is brutally assassinated, but also the comet (Sidus Iulium) is the physical manifestation of Caesar’s soul ascending to the heavens to join the gods.

Equally undefined is the concept of “discovery”, “creation”, “invention”, or “innovation”. For instance J. Robert Oppenheimer is widely known as the “Father of the Atomic Bomb” even though he is on many levels no such thing. If one had to ascribe the invention of nuclear weapons to one person that would be Leo Szilard. Of course since nuclear weapons were independently built in many countries, including Russia, there is no one person even if there were one person for America, which there is not.

If you are Marcus Vipsanius Agrippa you steal credit for the creation and inventions of your subordinates and Julias Caesar then steals most of the credit from you. However, ignoring the obsession with getting your name on things that are the product of thousands, the Via Agrippa roads have the real world utility of benefiting travelers. The real world utility of atomic power or a data center is much less clear.

Intelligence is also a much more insidious part of the educational system. Society can, if it tries hard enough, teach a subject without pitting students against each other by race or gender. Unfortunately, as long as tests are involved there is no intelligence neutral form of teaching.

The tests given to human students are both benchmaxxing and an inversion of benchmaxxing we’ll call divisionmaxxing. In divisionmaxxing the test is created to make similar test takers look the most different.

Its easy to divisionmax any test by using elements such as time limits, rote memorization, and artificial subject depth. Similarly divisionmaxxing your employees is a snap — just grant some of them house slave perks and keep the rest in the field.

On the question of whether people can be trained for the intelligence industry, AI, so don’t trust it, says probably not:

Here is the insidiousness of divisionmaxxing again. Education, professions, and sports are deliberately immune to training. Strolling around the block isn’t a sport, but power walking a marathon is.

Another undefined aspect in the intelligence industry is competition. If you “invent” something then it doesn’t seem fair that others should compete with you. In 2023 Musk says:

Nothing any of my companies have done has been to stifle competition. In fact, we’ve done the opposite. At Tesla, we have open-sourced our patents. Anyone can use our patents for free. How many companies do you know that have done that? Can you name one? I can’t. At SpaceX, we don’t use patents. So I mean, once in a while we’ll file a patent just so some patent troll doesn’t cause trouble. But we’re not stopping, we’ve done nothing anti-competitive. Some companies have done anti-competitive things. At Tesla, we have made our Supercharger system open access. We made our charger technology available for free to the other manufacturers. No walled garden. We could have put a wall up, but instead we invited them in.

In 2024 Biden put a 100% tariff on Chinese made electric vehicles. Even after donating $290 million to Trump and gaining unprecedented access to the president, those tariffs have never been reduced. However the unprecedented access did play a role in SpaceX getting billions and billions of government contracts for the golden dome and other benefits.

The money the government has put in all along to support Musk’s EVs could have been spent on trains. The benefits of a golden dome will likely never materialize in an era of ever cheaper drone warfare. The people who get rich in such companies will further corrode our political system with I deserve it divisionmaxxed beliefs.

Plus divisionmaxxing opens a door to erase gains made in tolerating other races and genders. Silicon Valley has a double tiered race and gender glass ceilings that is justified by claiming that only certain genders and races are good at technology.

Eventually women and other races will get across the moat of traditional race and gender bias but the most inner wall of the divisionmaxxing castle, intelligence, is a long way from being breached.

Demis Hassabis: I think science is the most important, and the scientific method is probably the greatest, invention humanity has ever come up with. I think science is what drives all progress. Modern civilisation comes from scientific and technical advances. I think the most important thing we could apply AI to is to advance science and things like medicine.

Hassabis doesn’t say it but it’s easy to take his comment as a divisionmaxxing waterfall:

Science is the most important => scientists are number one => engineers are number two => scientist and engineers should have greater status

Yet there are a lot of simple mistakes in the above quote. First off scientific method is nowhere near the greatest invention of humanity. Maybe it would make the top 10 in a list that includes language, clothing, carrier bags, fire, agriculture, domesticated animals, the wheel, ships, art, etc. — maybe not.

Second science does not drive all progress. The vast majority of human progress came before the invention of science. Arguably the vast majority of the world’s progress came before the invention of humans.

“Modern civilisation comes from scientific and technical advances.” — this is a tautology as modern = technologically advanced.

“I think the most important thing we could apply AI to is to advance science and things like medicine.” — also kind of a tautology since probably AI has the best shot at progress by curing diseases. However that’s not because diseases are are main problem but because AI probably can’t solve universal healthcare, the drug overdose crisis, poverty, etc.

Third the value of science is not a constant. As more and more is discovered the value of most inventions becomes less, not more, than the ones before it. So electrification sets an incredibly high bar for anything after it. Conversely the longer we put off changing a system born of colonialism, slavery, and unregulated manufacturing, the more that system, and not science, becomes the low hanging fruit.

An argument could be made for reversing the value waterfall:

Stability is most important => competent politicians are number one => ordinary workers and family are number two => tangible asset work, families, and policy makers should have greater status

That argument might fail in the face of AI actually curing diseases.

However AI driven disease cures may send us further and further down the divisionmaxxing rabbit hole if we simultaneously can’t provide universal health insurance, free college, etc.