What the Amazon Knows That Silicon Valley Doesn’t

A note before we begin: I work at the intersection of AI, climate, and Indigenous knowledge systems, which means I have professional and personal reasons to be optimistic about AI’s potential. I’ve also had experiences that have given me a direct relationship with non-linear, relational ways of knowing. I’ll name that bias upfront rather than bury it, and ask you to weigh what follows accordingly.

Across multiple major independent studies — Stanford’s AI Index, Pew Research, Ipsos, Gallup International — a consistent pattern emerges in global attitudes toward AI.

In China, 83% of people believe AI offers more benefits than drawbacks. In Indonesia, 80%. In Thailand, 77%. Vietnam and several other Southeast Asian nations exceed 90%.

Now to the other side. In the United States, that number is 39%. Canada, 40%. The Netherlands, 36%. France, Germany, and the United Kingdom all sit below 50%.

The countries that built the dominant AI systems, those that house OpenAI, Anthropic, Google DeepMind, Grok, are among the more likely to believe AI offers more harms than benefits. The builders are the most afraid.

That asymmetry stuns me and is worth sitting with. Not to dismiss the fear, it isn’t irrational, but to ask what it reveals.

What’s Driving Fear in Affluent Societies

The psychologist Daniel Kahneman spent decades documenting a pattern that holds across cultures and contexts: losses feel roughly twice as powerful as equivalent gains. We are wired to fear losing what we have more acutely than we hope for what we might gain.

Scale that psychology to a nation with accumulated wealth, established institutions, and entrenched labor markets. Disruption lands differently when you have more to protect. This isn’t irrationality, it’s a predictable response to a real asymmetry of stakes.

But loss aversion explains the fear in wealthy nations. It doesn’t explain the optimism elsewhere. And the optimism is the more interesting question.

Brazil Breaks the Simple Story

It would be tempting to draw a clean line: wealthy nations fear AI, developing nations embrace it. But Brazil breaks that frame usefully.

Brazil should be optimistic by straightforward development economics logic — large emerging nation, young population, significant appetite for technological leapfrogging. But Pew Research’s 2025 global survey places Brazil among the nations where concern about AI is highest. Roughly half of adults are more concerned than excited, clustering closer to the United States and Italy than to Indonesia and Thailand.

Brazil contains multitudes. São Paulo carries the labor anxieties of any sophisticated urban economy. A vast informal workforce faces genuine automation risk. Extreme inequality means the fears of the established middle class and the hopes of the economically marginalized point in entirely different directions.

The geographic frame, it turns out, is a proxy for something more precise. And finding that something more precise requires looking not at nations but at relationships — specifically, relationships to knowledge itself.

A Different Relationship to Knowledge

Chief Almir Narayamoga Suruí, leader of the Paiter-Suruí people of Rondônia, Brazil, has called AI “digital shamanism.” His people, and peoples like them, have been reading the intelligence of the Amazon basin for thousands of years; tracking ecosystem signals that no satellite has yet learned to detect, holding ecological knowledge across generational timeframes that no quarterly dataset can replicate.

When Chief Almir encounters AI, he doesn’t ask whether it will take his job. He asks whether it can help protect what his people already know — and whether the systems being built will treat that knowledge with the sovereignty it deserves.

That is a structurally different question than the ones being asked in San Francisco, Austin, Seattle, and Silicon Valley.

A Hypothesis Worth Considering

The communities most open to AI may tend to be communities whose relationship to knowledge is fundamentally relational. In these traditions, knowledge is not primarily something you accumulate and protect. It flows between generations, between species, between the living and the ancestral. It is held collectively and understood to be ongoing rather than fixed.

The communities most fearful of AI may tend to be communities whose relationship to knowledge is primarily extractive. Data is gathered, models are trained, intellectual property is protected. Knowledge is a resource. It can be owned, lost, displaced.

When AI arrives into a relational knowledge tradition, the question it raises is generative: can this tool help us protect and transmit what we already know?

When AI arrives into an extractive knowledge tradition, the question is defensive: will this tool take what we have built?

I want to be careful not to overstate this. The data shows geographic and economic correlations. The epistemological explanation I’m offering is one interpretation of what might lie underneath those correlations, not the only possible one. A thoughtful skeptic would rightly point out other variables: media environments, political cultures, specific labor market conditions. Those explanations are also valid.

But the pattern is consistent enough, and the framing resonant enough with what practitioners in these spaces actually report, that it seems worth naming honestly and examining seriously.

The Question Underneath the Question

Here is what I find genuinely unresolved, and worth sitting with regardless of where you land on the interpretation above. The communities with the most intact relational relationships to knowledge are also the communities whose ancestral intelligence is most at risk of being absorbed into AI systems that will not give it back. The very openness that may come from a relational epistemology is also what makes these communities vulnerable to technology designed from extractive premises.

Which is why how we build AI may matter as much as what we build.

What would it look like to genuinely begin from the relational premise, not as a values statement but as a technical and governance commitment? To treat Indigenous ecological knowledge as a co-equal intelligence stream, with sovereignty encoded at the architectural level rather than promised in a policy document?

Those aren’t questions I can answer here. But they are the questions that the most interesting conversations at the intersection of AI, climate, and Indigenous governance are beginning to seriously engage.

If Chief Almir’s framing resonates with you… if the question of what the Amazon knows, and whether AI might help protect rather than extract it, seems worth serious attention… then the conversation is happening, and you’re invited into it. On April 21 in San Francisco — The AI for Nature (mini) Summit: https://luma.com/xm0nve4v

Scott Broomfield is Founder and Executive Director of the Pachamaya Foundation, working at the intersection of AI, climate, and Indigenous knowledge systems. Previously he was the CFO of Pachama and 1mind.