Won’t someone think of the historians?
Many of you will have seen the Microsoft research that was circulated online, listing vocations that were most and least vulnerable to AI. I read the paper, and to be fair, it has a proper methodology, it’s not just some opinion based metric. The researchers generated a score based on whether people in a particular occupation used AI, and how widely. They state: “The score captures whether AI is being used (with sufficient activity share) for the work activities of an occupation and whether that usage tends to be successful (completion rate) and cover a moderate share of the work activity (scope)”. This seems like a reasonable way to approach the problem, although it is not without issue, as they admit “decomposing an occupation into its work activities, while standard practice in the literature, does not provide a complete representation of every occupation: the connecting glue between tasks also contributes to the value of work.” That is, only some aspects of an occupation are captured by this work.
Before we get on to the wailing, it’s worth noting that the authors add some nuance which is missing in most of the online discussion (I know, who would’ve guessed?). They are careful to point out that this does not mean total usage by any occupation, and so talk of replacement is premature: “Our data do not indicate that AI is performing all of the work activities of any one occupation. … There are definitely some
occupations for which many—perhaps even most—work activities have some overlap with demonstrated AI capabilities. But even when there is overlap, the task completion rate is not 100% and the scope of impact is usually moderate.”
It is that last element of scope that I think is important – the range of work undertaken. Like many people, seeing Historians ranked highly at risk caused some alarm when I saw it. This is not a criticism of the research, but rather of how some people will interpret it and what that means. As I suggested before, the fetishisation of summaries is problematic in general, but in history in particular. If the AI version of history becomes the default then the adage that the victor writes history can be rephrased as the powerful rewrite history.
Let’s take one example because we can see it playing out right now, but you can imagine this for any aspect that is deemed unacceptable to authoritarians. It does not require a dystopian novelist to imagine a near-future where an AI version of history completely ignores, or rather is unaware of, trans people serving in the military. When a government is actively banning trans people from serving and denying them early retirement, and removing them from websites then their complete removal from all historical records is the next step. The White House order could probably be used to implement this already.
An AI historian will simply regurgitate the data it has. And this gets at the function of historians. It is not to simply regurgitate dates of events, but to interpret the present, to reinterpret the past, to question accepted narratives, and to bring to light forgotten or lost accounts. Historians are, in their way, radical. One of the key components to the current far-right surge is the use of narrative – immigrants are causing crime, trans people are attacking kids, the country is broken – you know the sort of thing. These all have emotional power regardless of the actual truth (eg that crime rates are on the decline). Everything was better in the past and populists can return you there. One of the roles of historians is to challenge these narratives and offer interpretations of the past that give voice to the disempowered.
An AI version of the past becomes a tool for the powerful and the dominant narratives. So, while I’m sure historians will use AI as part of their toolkit, we shouldn’t confuse this with AI replacing historians.
