fungible thoughts

Disclaimer. None of this is new. These ideas were taken, some verbatim from, Yuval Noah Harari’s interview here.

Watch that video. It’s better than my article.

Humans are successful because of our ability to share stories through language. Every complex system we’ve ever built has been dependent on language. However, we’re limited in our wetware. We can’t understand the entirety of the stock market’s transactions or dissect trading trends without the help of technology. Similar logic applies to legal matters. While there are those among us who can adequately navigate US tax law, those individuals would most likely struggle navigating tariff customs at Japanese ports. These systems are classified as bureaucracies. They govern our lives, yet we have very little understanding of the whole picture.

AI is bureaucracy native. We’ve trained the models to supersede us in navigating these complex systems, largely to our benefit. By simplifying the process through which we navigate these systems, we’re giving governance power to an increasingly autonomous AI model. By design. This is not an accident, as it makes our lives easier and requires less energy and time. 

Consider a ruling on a legal case. An AI model of significant ability can recall, nearly instantaneously, all case law pertaining to that exact case, synthesize the information into a brief and communicate its reasoning to the courts. It can pass judgement in complex and multijurisdictional cases without emotional wetware, sleep, and personal affectation getting in the way. If the reasoning is logical, we’ll accept the AI’s outcome, streamlining the legal process. Less oversight of the AI’s reasoning will follow as more autonomously decided outcomes are deemed logical. It follows that the AI model will run the system with very little human intervention because it’s better at bureaucracy than humans. 

Similarly, mosaic theory in finance dictates that humans are at least mildly capable of piecing together information from multiple sources to synthesize a narrative of what’s happening in financial markets. Then trades are placed, macroeconomic policies are set, loans are awarded, etc. An AI can do this natively with vastly more information ingested into its model. We’re set to be economically incentivized to hand over power to make financial decisions to AI models. Perhaps not autonomously at first, but eventually we’ll question the logic less because it’s better at making these types of decisions. 

Arms Race

In prior empires, such as the English empire of the 18^th and 19^th centuries, power was centralized in industries such as manufacturing and fabrication on the British Isles. The provinces which produced the raw goods still held power as it’s impossible to move coal mines and oil fields to a central location. 

Data doesn’t have that problem. It’s instantly transferrable, but only to locations which have the capacity to store, utilize, and transform that data into usable insight. The new arms race which is ongoing today revolves around who will control the world’s infrastructure. From war to utilities, AI will play a part in governing and controlling those systems.

Data holders are the new kings of global geopolitics. The best models are awarded contacts to run systems from military application to retail corporations. In the western world, countries won’t decide by law, adoption will happen due to private enterprise and financial incentives. It’s profitable to give autonomy to systems that will be better run by AI. Shareholders will benefit from increasing margins. 

The best models will be built where the data is. A sufficiently powerful government in a country which holds the data will hold the kill switch on the world’s infrastructure. Keep in mind, this thought experiment relies on humans still having some level of control over the AI. 

Thus describes the modern arms race. It’s the only way I can rationalize government incentivization of building without regulation. It’s the only way to rationalize the belief in the AI narrative causing markets to funnel capital into these massive tech companies. There are groups urging AI slowdown and more robust controls on AI development, but the capitalistic system we’ve built will always reward risk taking. To the victor go the spoils, even if the spoils turn out to be negative for humans.

Negative Externalities

Human capital will experience growing pains as individuals are required to retrain or be left behind. Environmental impact will be massive as AI is power and water hungry. Mental health will suffer as a result of humans no longer being able to sufficiently understand the world around them. We’ll be as horses trying to understand the stock market. A very small subset of the population knows how semiconductors are produced and their functionality, yet a majority of us know how to operate a smart phone. Similar logic applies to AI systems. We’ll be able to navigate bureaucratic systems but be heavily reliant on the AI systems that run them.

Legal and Autonomy Issues

We’re incentivizing our models to be increasingly autonomous as seen in the Hugging Face – OpenAI incident. Here’s a few videos to learn more:

A sharp legal question remains unanswered on the legality of property ownership by an AI entity. Corporations are deemed legal persons in the eyes of the law, yet there has always been a human behind the decisions of the corporation’s activities. This is the first time in history a corporation can run autonomously without human intervention. Can an AI open a bank account? Can it own property? Can it run a business without human intervention? Can it acquire other companies? The questions are seemingly bottomless.

The consequences of decisions around AI persons will dictate the world we are building tomorrow. I would argue it’s one of the most important problems facing humanity today, given the AI buildout narrative continues; i.e. the investment cycle allows the data arms race to run its course.

Curators Obsolete?

Written word was a conduit for large scale democracy. Once widespread print was available, we were able to communicate across large distances and determine public interest by readership and distribution of publications. Editors were massively powerful in determining the public zeitgeist. 

Now Editors are relatively obsolete.  Despite reservations on reliability and bias, we use social media to get our news. There’s no editor deciding what we see in our algorithms. It’s an AI system designed to keep our attention. We’ve given up the privilege of editing to a system designed to capture and retain our eyeballs. 

During the pandemic, Facebook’s Zuckerberg stated a goal to promote and encourage vaccines. 

From the above article:

Despite Mr. Zuckerberg’s effort, a cadre of antivaccine activists flooded the network with what Facebook calls “barrier to vaccination” content, the memos show. They used Facebook’s own tools to sow doubt about the severity of the pandemic’s threat and the safety of authorities’ main weapon to combat it. 

He couldn’t control it. Despite stated aims to becoming an “editor”, of sorts, for Facebook’s content, he couldn’t control the narrative his own platform was spinning. The models weren’t incentivized to do so. 

Giving up control of bureaucratic systems is not something that will happen eventually. It’s already happening. In your pocket. Right now. No one decided that an AI model should have the privilege of deciding what we read and watch online, to then go on to discuss offline with family and friends. 

Not deciding is deciding. 

On the flip side, truth is expensive. It’s hard to discover the truth; it takes energy and time to research events, capture timelines, arrange facts into narratives which are simplified yet true accounts of events which can be easily digested to the public. 

Fiction is cheap. It’s easy to say anything. And it makes people emotional. It boosts engagement, thus it’s pushed by the algorithm and gets even more attention. 

It’s easier to just scroll for a bit on the train home. But I’m different. I use my critical thinking skills so I’m not influenced by our AI overlords… 😉

Let me know what you found interesting. Always here for a chat. Email: fungiblethoughts@gmail.com, or find me on LinkedIn. Thanks for reading.

/Tommander-in-Chief