Here’s a fun google search for kids: “Which country has the most tigers”
The answer is India, which is what Google tells us without even a link. And then there’s a gazillion links that back that up. India, India, India. There’s about 3,000 tigers in India.
But there’s a big fat caveat, which is easier to see in the WorldAtlas version: “Countries With The Greatest Number Of Wild Tigers”
Wild. Tigers.
Ok, so here’s another fun google search:
“Which country has the most tigers in captivity”
And suddenly you get questions in Google like “are there more tigers in the US than in the wild?” and a whole bunch of interesting articles painting an entirely different narrative. By this story, there are north of 5,000 tigers in the US, mostly in Texas and Florida, and they easily outnumber the total population of wild tigers in the rest of the world.
So which is right?
What’s so interesting is how these questions are framed. It’s obvious what people mean when they ask a question like “which country has the most tigers”. We have this Platonic idea of a tiger in our heads, and it involves a gorgeous, powerful orange and black striped tiger living on it’s own in jungles or the wild or wherever. Clearly, this is what we mean. What’s so incredible is that our narrative ideal of a tiger is baked into the technology of Google. Assumptions are made based on searches and page ranks exactly what we mean when we ask about tigers. It’s so subtle that it can be really difficult to see.
Balaji has devised a framework that let’s us think about this sort of distinction. He places truth on a spectrum with political truths being on one end and technical truths being on the other. Political truths are constructed by consensus - for example, enough people agree on the border of a country and that’s the border. This is different from the technical truth defining the diameter of the earth. In Balaji’s words: “Political truths are true if everyone believes them to be true — money, status, borders. You can change these by rewriting facts in people’s brains: what is a dollar worth, who is the president, where is the border? This is what our establishment is set up to manipulate.”
Back to our tigers: we’ve got two narratives: There’s 4,000 wild tigers in the world with about 3/4 of them living in India. And we’ve got over 5,000 living in Texas and other states. What is true? Where are the major populations of tigers?
It takes a little more googling to build up a more technically true picture of tiger populations. There’s a long list of population claims that all say the same thing: 5,000 or 5,000-7,000 or 7,000. They’re all from news sources and all seem to be repeating each other, creating a political truth that’s demonstrated by the number of headlines. But there’s also been several studies done on US populations to try to get to the bottom of the Tiger King view of the world. These studies conclude that the overall count of private tiger ownership in the US is probably not as high as the headlines and is better estimated in the 800-3,000 range.
On the other hand, we’ve missed another huge population of tigers in China and Southeast Asia. There are over 200 tiger farms in Asia, setup as tourist and breeding centers. Harbin Tiger Park alone claims to have 800 animals and the total estimate by EIA is around 8,000 animals.
It certainly seems true that there are more tigers in the world in captivity than in the wild. It also seems true to say that there are three major population centers of tigers: China/Southeast Asia (captive), India (wild), and the US (captive). This, finally, seems like a reasonable - but still probabilistic - technical truth about the population of tigers. There is a technical truth of the population of tigers, but it’s unknowable. All we’re left with is political truths. How far we dig determines how close our approximation can get.
It wasn’t hard to dig for all this information. It took the ability to search and follow some different threads in different sources. But it’s also true that the initial headlines that Google gave me were incomplete. It’s not that they weren’t accurate; rather, there were assumptions involved that meant these were political truths. By consensus, we’ve decided that what we want to count when we consider “tiger population” is wild tigers. By agreement, media outlets have agreed to point to a figure of 5,000 in a post-Tiger King world as the appropriate measure of big cats in the US. I don’t know about you, but I expect this of the media. And unfortunately, because of Gell-Mann amnesia, we’re all apt to believe in political truths far more than we’d like to believe. You could even say that our ability to correctly read a newspaper is itself a political truth. But a Google search? We think of Google as apolitical, unbiased and algorithmic, but of course it isn’t.
Any system that builds upon the collective data of humanity will enshrine all of the biases, assumptions, and mistakes of ourselves. As more and more of our systems are built on ML/AI ideas, these problems become more obvious. The people building these systems think deeply about problems of model training and data sanitation. These concerns will improve over time, but what I’m more curious about is how we intuit and interact with the outputs of these systems. Just as Gell-Mann amnesia makes us ignorant of the mistakes of newspapers even when we know they exist, so to does our willingness to go along with AI systems blind us from their issues.
Balaji’s framework let’s us more weary of the problem. Some of the most prominent hot-button issues are debates about a kind of truth. The transgender movement is a disagreement between political and technical truth: one side screams that biology is immutable while the other screams that gender is consensus. Crypto, climate, and racism can all likewise be seen through both political and technical truth arguments.
There are far more political truths in our world than we would care to admit. Including those that appear as technical truths.