Bureaucracies & Artificial Intelligence
One of AI's biggest challenges may be bureaucracies. Also, it may not be easier to get your fishing license.
Depending on ones experience with them, bureaucracies are either one of humanity’s greatest innovations, or one of our worst. They exist in all corporations and governments of all political systems around the world. Oddly enough, while we think of the headwinds Artificial Intelligence (AI) faces such as regulations, laws and the massive energy requirements, it may be bureaucracy that becomes a huge headwind. Including your ability to get a fishing license.
Bureaucracies have been around for a rather long period of time, dating back to when humans started to live in larger communities as we became agrarian. As we built cities, we needed administrations to run them. Both the Romans and the Egyptians were stellar bureaucracies. It is believed the first bureaucracy came out of ancient Sumer, around 3,500 B.C. Over time, they have evolved to become a core element of all societies.
While we might like to think that as we learned to write and record things, we first wrote lovely sonnets, poems and such. The reality is a bit more mundane. We kept records of things in warehouses and administrations recorded the minutiae of debts owed, fines and so on. Fascinating to archeologists and historians, a cure for insomnia for most. In modern societies, much the same happens, except we use Information Technologies and massive databases. As is said, the more something changes, the more it stays the same.
The word bureaucracy comes from the French language around the mid 18th century. It combines the French word bureau (desk or office) and the Greek word kratos — rule or political power. The French are quite brilliant at bureaucracy, so it makes sense they’d come up with a word for very old institutions. And turn it into something rather spectacularly confusing and complicated. India probably has the world’s largest bureaucracies.
Artificial Intelligence in Bureaucracy
AI tools from Machine Learning to Natural Language Processing, Neural Networks, Cognitive Computing have been used for around twenty years, some longer, in various bureaucracies, which include corporations, especially large ones. Now, Generative AI (LLMs, RAG etc.) are seeping into some government bureaucracies, which is where I will focus for the sake of this article.
Bureaucracies have used AI tools for policing (law enforcement is the violence arm of bureaucracies), judicial systems (where there’ve been lots of disasters such as racial and gender biases), city planning (some very good use cases), transportation systems and economic modeling among others.
These applications of various AI tools have faced minimal pushback from bureaucrats themselves. They’ve not really threatened jobs and have often proven to enhance them. And will continue to do so. Over time, bureaucracies, hopefully, will correct the wrongs of early applications. It won’t be a utopia, it won’t be a dystopia.
Generative AI (GAI) however, may prove to be a bit of a sticky wicket for bureaucrats. Including getting fishing licenses.
Bureaucrats Vs. Generative AI
If I were to come up with a conspiracy theory around government bureaucracies it would be that they all have a Department of Creating Confusion, which would probably have a rather mundane sounding name like Department of Internal Efficiencies. Their sole job being to figure out how to keep the general public utterly confused and provide long PowerPoint decks for training sessions on how to play a shell game with forms when dealing with the public.
This department would probably love GAI, if it could be used to create ever more complex forms and documents. As AI social agents come along and workers are assigned their own agent, then their agent might duke it out with the Department of Creating Confusion’s AI agents. Thus further delaying the issuance of fishing licenses.
Bureaucracies are very hierarchical and deeply political, both in culture and in nature. As AI agents are deployed to workers and management, these agents will learn from the behaviours and attitudes of workers and management. It is a feedback loop. We train them, they train us and on it goes.
Bureaucrats like to send lots of emails, write papers, create very large PowerPoint decks and have lots of internal finger pointing fights. Perhaps GAI would be very useful here. Instead of the humans doing all this sort of mundane stuff, the AI agents could do it. And have digital finger pointing fights too. It will probably still be just as confusing to get your fishing license though.
Humour aside, there are some solid use cases for GAI in bureaucracies. That may also be the place where GAI runs into some bumps. Many bureaucrats are also unionized. If an AI agent of a worker makes a mistake and issues the wrong fishing license that lets a fisher use a trawl net in a lake to take out all the fish, who gets the blame? The worker or the AI agent? If it’s the AI agent, then that AI agent would have to have its algorithms changed. And be re-trained. Lots of work, time and money.
As the general public becomes concerned with AI agents and how they are used and act, they will put pressure on the elected politicians. Who will then put pressure on the bureaucracy to come up with fixes. Of which some may become laws and regulations.
So it may well be not law makers, but bureaucrats that decide the ultimate fate of GAI and other AI tools. Or maybe their AI agents will figure that out? One wonders, will AI agents create massive slide decks and compete with one another to create more complicated forms and processes?
Perhaps citizens will have their own AI agents who can then argue with the bureaucracies AI agents? I think it’s safe to say, if we can figure out how to make AI agents make bureaucracy even more mysterious and complex, history shows we will. The introduction of IT into bureaucracies has made websites more complicated, email strings longer, the printing of paper more, not less and entrenched fax machines. I’m going fishing.