Transcript
Welcome to Iconoclast Insights. I'm André Daus, and today I want to talk about something that makes most people uncomfortable: the impossibility of getting everyone's consent, and why pretending otherwise might be the most dangerous thing we can do.
Last week, I attended a meetup with AI enthusiasts. The main speaker talked about founding a nonprofit focused on Artificial General Intelligence alignment and ethics. Three people. Just three. And after a year, they still couldn't agree on what ethical AI means. Then, in the same breath, she insisted that humankind needs to collectively define ethics for artificial intelligence before we proceed.
I sat there thinking: if three motivated people with similar backgrounds can't reach agreement in twelve months, how exactly are we supposed to get seven billion people to consensus? And more importantly—why do we keep pretending this is even possible?
Let's be honest about what's happening here. When we talk about "getting consent" or "defining ethics for AI," we're engaging in a particular kind of theater. It's a performance that makes us feel responsible, thoughtful, and careful. But it's also a performance that conveniently ignores reality. And here's the reality: ethics, fairness, and what's "right" are not universal truths waiting to be discovered. They're deeply subjective, shaped by culture, circumstance, and—most importantly—need.
I once heard Konstantin Kisin speak about climate change at the Oxford Union. He spoke about woke culture, but we do not want to go down that rabbit hole now. Anyway, he explained why many Europeans struggle to understand why developing nations aren't prioritizing environmental action. His point was simple but devastating: for much of the world, the primary problem isn't saving the planet. It's not starving today. When someone in a developing country has to choose between feeding their family and reducing carbon emissions, they choose their family. Every single time. Not because they don't care about the planet, but because survival takes precedence over philosophy.
This isn't about food versus ethics. It's about understanding that people's priorities are determined by their circumstances. And here's the uncomfortable truth: ethics discussions are a luxury good. They're something people engage in after their basic needs are met.
So when we talk about getting global consent on AI ethics, we're really saying: "Let's ask everyone to agree on something, including people whose daily reality we don't understand, whose priorities we don't share, and whose voices we often can't even hear." About one-third of the world's population doesn't have access to AI yet. They can't participate in this discussion even if they wanted to. But somehow, we're supposed to wait for their consent before proceeding?
Let me give you a closer example. In Germany, we had the Pirate Party—a political movement built on radical democratic principles. Every member who wanted to speak could speak. Perfect democracy, right? The results were actually quite good. The problem was it took forever to reach any conclusion. Party meetings became endless marathons of discussion. Everyone got their say, everyone's voice mattered, and eventually the party discussed itself into irrelevance. I'm not even sure they exist anymore in any meaningful way. This is what happens when you prioritize universal participation over action. You don't get better decisions. You get paralysis.
Now, living in Germany, I see this pattern everywhere. We love to discuss. We pride ourselves on our "Gründlichkeit"—our thoroughness. We created data privacy regulations that are now standard across the EU. My website's privacy policy is over twenty-one thousand words long. Twenty-one thousand words explaining how I handle visitor data. The people who wrote these regulations don't seem to understand how the internet actually works, but they were very thorough in creating rules about it.
We're so proud of this carefulness that we assume everyone else should adopt it too. Meanwhile, German startups receive a fraction of the investment that comparable US startups get. When it comes to AI and business innovation, we're not leading—we're trailing. We demand that AI be completely transparent, as if neural networks operate on simple if-then-else logic. We never stop discussing, never stop requiring one more round of consensus, one more compromise.
And there's that word: compromise. In Germany, we worship compromise. It's seen as the mature, responsible approach. But ever since I read "Never Split the Difference," I shudder every time someone suggests a compromise. Because compromise usually means everyone gets less than what they need, and no one gets what they actually wanted. It's the illusion of agreement covering up fundamental disagreements that haven't been resolved.
Don't misunderstand me. I'm not saying we should rush recklessly into artificial general intelligence. I'm not advocating for moving fast and breaking things when "things" might include humanity's future. Thinking before acting is crucial. Creating guidelines before starting something massive is smart.
But—and this is critical—we can also discuss everything to death before it even starts. Germans excel at this. We can think our way into complete inaction. A friend once told me: "It's better if some people start toward the goal in a sports car than waiting for everyone to board the bus." His point was that when you're trying to reach a new destination, you need a small group moving quickly, testing the path, discovering if the goal is even the right one. You don't wait for everyone.
This works in organizations where leadership can influence every employee. But when we're talking about humankind, there is no leadership structure. There's no control. Every human being acts based on different purposes, different needs, different worldviews. And frankly, that's a good thing. We shouldn't want centralized control over seven billion people. But it also means we need to abandon the fantasy of universal consent.
So what do we do instead? Here's what I think: we need to acknowledge that complexity cannot be solved, only managed. We can't wait for perfect agreement. We can't achieve universal consent. What we can do is be pragmatic.
There's a German saying: "auch mal fünf gerade sein lassen." Literally, it means "sometimes let five be an even number." In practice, it means: be pragmatic, don't be overly precise, accept that perfection isn't always possible or necessary.
Instead of seeking impossible consensus, we need transparent experimentation with clear boundaries. Small groups moving forward, yes, but with visibility. Not hidden development that suddenly presents the world with a fait accompli, but open progress that others can observe, critique, and learn from.
We need to accept that different cultures, different nations, different groups will have different approaches. That's not a problem to solve—that's reality to manage. The goal isn't agreement. The goal is coexistence of different approaches with mechanisms to prevent catastrophic outcomes.
And we need to be honest about who's actually in the room when we talk about ethics. If your ethics discussion doesn't include voices from people facing food insecurity, lack of internet access, or fundamentally different worldviews, you're not having an ethics discussion. You're having a preference-sharing session among people who already largely agree.
The danger isn't that we'll move forward with AI without perfect consensus. The danger is that we'll freeze in endless discussion while others move forward anyway, just without our input, without our perspectives, without our concerns. The people most committed to getting everyone's consent might be the ones who end up with no influence at all.
So here's my question for you: What if the most ethical thing we can do is acknowledge that universal ethics are impossible? What if accepting our differences, working transparently within smaller groups, and managing complexity rather than trying to solve it is actually the pragmatic path forward?
Sometimes five does need to be an even number. Sometimes action with imperfect information beats perfect planning that never leads to action. Think about it.
This has been Iconoclast Insights. Thanks for listening and see you next time.
