Deconstructing Dutch Colonial Myths in Brazil

I had the pleasure to present the work we did with Sins Beneath the Equator at MuseTech, a recurring one day conference on Museums and Technology, this year hosted in Birmingham.

My presentation specifically focussed on using AI features to shorten the time needed for the production and delivery of content. The slide deck is below, an abridged version of the narrative follows.

In 2015, my not-yet-wife and I visited the lovely town of Olinda, in the Brazilian Northeast. Early in the 17the century, Olinda was the capital of the Portuguese captaincy of Pernambuco. At the start of the Dutch 24-year reign in the north of Brazil, they burned down the city, shifting the capital to Recife.

The city was rebuilt, and is lovely. We were strolling through town, admiring the facades, not saying anything, when a man came around the corner, and started talking to us.

We were obviously tourists, but he could not know my nationality is Dutch. Yet, one of the very first things he said was that “Oh, if only the Dutch had stayed, Brazil would have been so much better, today”.

Well. Knowing what the Dutch did, after Brazil, in what is now Suriname and Indonesia, this is, of course, bullshit.

Two years ago, I had the opportunity to attract funding for a project that would highlight untold stories of the colonial past. I jumped at the opportunity, found partners, and started working on Sins Beneath the Equator, putting together the mostly untold story, at least for the general public, of the Dutch Colonial past in Brazil.

We had a diverse team, but, with a limited budget, sought to speed up the process where possible, as we set out to produce 100 location-based podcasts in three languages, with stories geolocated on 5 continents.

The AI tools that now, seemingly, everyone is using, were very useful, but not perfect. Our findings:

• Machine translations were excellent to start with, but needed to be edited by a (near-)native speaker, yet still saving lots of time.
  Deepl and ChatGPT outperformed Google Translate.
• Machine generated illustrations by Midjourney produced excellent results.
• Machine generated audio from ElevenLabs was more than good enough, but a discerning ear would pick up the artificial undertones.
• Scaffolding articles, constructing the high-level narrative, with ChatGPT, as a basis for actually writing articles, was extremely helpful.
• Having ChatGPT do original research saved large amounts of time, but all produced claims needed to be individually fact checked, because of regular hallucinations.

However, though there is a clear advantage to using AI tools, there are also huge problems, which cannot be ignored and, one way or another, will have to be addressed at some point.

• The cost to the user is massively underpriced; OpenAI lost 12 billion USD in Q32024.
• All LLMs exist as a consequence of blatant copyright theft; Facebook trained AI on pirated material.
• All mainstream AI platforms use excessive amounts of energy; Generating an image uses as much energy as charging a phone.
• Big Tech builds an economy for the few, not the many. This results in:
  • Concentration of power
  • Wage collapse
  • Unforgiving data extraction
  • Hidden bias, potential uncontrollable fraud
  • Epistemic collapse (what is true?) and information chaos
  • Erosion of human creativity and agency
  • Outsourcing of death and destruction
  • Accelerated inequality

The British and Dutch chartered companies were the first implementations of capitalism. Big Tech, in the time of AI, are the most recent. The politics of exploitation and abuse are the same.

The destruction of Olinda served Dutch capital, but what lessons do we learn?