Six month ago I wrote a post called Ghost Shirts, Guilds, and Generative AI, which started with the infamous “Pause Giant AI Experiments” letter. In my post I looked at how groups historically tried to block or reverse change that the general public experienced as inevitable. But there are always options. I wrote about options for the situations where you don’t want that inevitable change:
“You could fight it indirectly and delay how fast the change happens. In that case, you will quietly subvert the system.
“You could fight it directly, even though you will probably lose. In that case, you are fighting for honor.
“Or, through a combination of luck and foresight you could build a system that shields you from the inevitable change taking over your corner or the world. In that case, you need to build and defend a boundary.”
Authors of the letter requested a pause of at least six-months in training AI systems more powerful than GPT-4. It’s now six months after their letter, so I thought it was time for another look at the topic.
Melos
One of the many things the world is dealing with now is the inevitability of AI having a big impact on the way we live. We could compare that inevitability to a historical conflict between two groups: one weak and one strong. But a commonality of weaker groups in history is their isolation — their physical inability or cultural unwillingness to learn from, adapt to, or combine with other groups. What can we learn from this?
We’re not used to thinking about such situations when it comes to ourselves in the present. We more often encounter something historical like the Melos Dialogue from the Spartan War.
Melos was neutral in the Spartan War, but was the only significant island in the Aegean Sea that Athens didn’t control. Athens’ negotiators eventually arrived to demand Melos take sides and pay tribute. What’s interesting about the dialogue and the reason we still study it today is that Melos’ negotiators — the weaker side — declined. They chose honor, which came with the risk of annihilation.
Athens then set a siege. Melos eventually surrendered, the Athenians executed the island’s adult men, and enslaved everyone else. To the Athenians, it was just common sense. From the dialogue:
Athenians: “For ourselves, we shall not trouble you with specious pretenses… and in return we hope that you… will aim at what is feasible… since you know as well as we do that right, as the world goes, is only in question between equals in power, while the strong do what they can and the weak suffer what they must.”
Who Are the Strong? Who Are the Weak?
Dealing with AI, some propose bans or regulatory exclusions in certain arenas. Such actions come from the currently strong (notably including those in the entertainment industry as well as government). But strength can be a local measure. How do these decisions impact what will come later?
As a way to learn, I chose from the long list of isolated peoples encountering an entering force.
Here’s a short list of some of past encounters between isolated groups and entering groups. In these examples, the contacted isolated groups were all based on islands.
- Han Chinese – indigenous Taiwanese. Hundreds of years ago, Chinese migrations and expansions took them to Taiwan, resulting in cultural exchanges, conflicts, and the assimilation of some indigenous groups. The later 1949 Nationalist retreat to Taiwan from mainland China led to a white terror, the 2-28 massacre, and the still common control of many government and media positions.
- Maori – Moriori. The Moriori arrived on the Chatham Islands around the year 1400 and lived in isolation for 400 years. In 1835, two Maori groups from New Zealand invaded the Chatham Islands. While the Moriori met them peacefully, the Maori killed and enslaved the Moriori.
- Spanish – Canary Islands. In the early 1400s the Spanish enslaved the island’s Guanche people before intermarrying with them, forming the modern Canarian culture.
- French – New Caledonia. Both British and French explorers visited New Caledonia in the late 1700s, but France didn’t take possession of the islands until 1853, when it established a penal colony. France then restructured existing tribal leadership and imposed forced labor and curfews. The French suppressed multiple uprisings and deported or killed the rebels.
- Dutch – Banda Island. In 1621, Dutch colonists conquered and nearly wiped out the local Bandanese people in order to gain a monopoly on nutmeg and mace.
Distinct from above, the following groups are not physically isolated on islands, but still remained separate from their surroundings. The examples below are of peoples who chose to invest in preserving their way of life (defending a boundary), even as they were surrounded by larger groups or social change.
The following mostly use the boundary defense tactic, whether intentionally through religion and culture or through careful consideration of change.
- Amish in the US and elsewhere. The Amish will adopt certain new technologies, but carefully and after considering whether it is good for the group. Amish communities have bought cheap farmland and established new colonies to support their growing population. Population: 360K.
- Coptic Christians in Egypt. This group preserved their language, religious practices, and traditions while living within a predominantly Muslim society. Population: 5 – 20M.
- Zoroastrians (Parsis) in India. This group is descended from Persians who migrated to India centuries ago. They maintained their Zoroastrian faith, rituals, and traditions. Population: 6K.
- Tribes in the Amazon rainforest, including the Yanomami, Kayapo, and Ashaninka. These groups maintained their languages and independence (though at the whim of loggers, miners, and the Brazilian government). Population: in the thousands
- Saami people of Kola Peninsula in Russia. This group maintained their language, reindeer herding practices, and culture while living within larger Russian communities. Population: 50 – 100K.
As subversion and honor examples, in Ghost Shirts, Guilds, and Generative AI I mentioned examples of dancing and flagellant “epidemics” in Europe and the US. These epidemics served a human need, even if they had minimal results. The other example of the iron guild resisting the steel industry was a more modern social example of people trying to preserve a way of life to which they had grown accustomed.
These were actions of the weak trying to subvert or fight the strong, rather than defend a boundary. And those examples failed.
Back to Melos. As a very partial recovery after mass execution and enslavement, Athens’ loss in the Spartan War led to Spartan military leader Lysander returning the island of Melos to the previous inhabitants that could be identified.
Executive Order on AI
On October 30, 2023 the Biden administration issued an “Executive Order on Safe, Secure, and Trustworthy Artificial Intelligence.”
It attempts at “seizing the promise and managing the risks” of AI. At least as stated it seems like a mix of defending a border, subverting to delay change, and fighting to stop change. A few of the sections and my interpretation of their purpose:
- Subvert / Delay. “Develop standards, tools, and tests to help ensure that AI systems are safe, secure, and trustworthy.” I really wonder how we could accomplish this or whether the statement even makes sense.
- Fight. “Protect against the risks of using AI to engineer dangerous biological materials…” Another way to look at this would be to intentionally use AI for those purposes but with the aim of then guarding against such materials.
- Fight. “Protect Americans from AI-enabled fraud and deception…” Ironically, this was the topic of my first on unintended consequences, which focused on voice AI.
- Subvert / Delay. “Evaluate how agencies collect and use commercially available information—including information they procure from data brokers…” Interesting to include this and I’m happy to see any oversight of the the data broker industry. I earlier wrote about the “Techno Richelieu Effect,” where AI identification of behavior produces statistical guilt. And since so much data is tracked, everyone has that guilt.
- Defend a Border. “Ensure fairness throughout the criminal justice system by developing best practices on the use of AI in sentencing, parole and probation, pretrial release and detention, risk assessments, surveillance, crime forecasting and predictive policing, and forensic analysis.” The inclusion of predictive policing was the one that stood out for me. How will crime prevention change because of AI? But my read on recent problems using AI for suspect identification is that they can fail more because of human laziness than anything else. In the famous Robert Julian-Borchak Williams case from 2020, AI analysis of video of a crime identified the wrong person, but he was actually arrested because of a poorly run police process rather than bad facial recognition tech.
- Subvert / Delay. “Develop principles and best practices to mitigate the harms and maximize the benefits of AI for workers by addressing job displacement” is a point on slowness, almost like an Amish approach. Or would this simply stifle an industry?
- Defend a Border. “Use existing authorities to expand the ability of highly skilled immigrants and nonimmigrants with expertise in critical areas to study, stay, and work in the United States…” and “Accelerate the rapid hiring of AI professionals as part of a government-wide AI talent surge”
Depending on how you see yourself and how comfortable you are with AI-related change, you might choose or push for different choices. To me, this is a strong encountering weak situation. The world won’t go back to the way things were pre-AI but we can try to maneuver through the change.
Consider
- I previously wrote about historical choices to isolate the body and isolate the mind. The people making those choices looked at the outside world as something to temporarily leave, for reasons of pandemics or social decline.
- The AI debate deals with issues that permanent change, of no going back. There may be, like the Melos example, situations where the strong give the weak back some of what they lost or intentionally want to preserve “historical colonies,” whether out of cultural interest or embarrassment.