EP 2: Who Controls the AI?

AI translation — original text in German.

Transcript

Question: Can an AI become dangerous? To me? To other people, to humanity as a whole?

Answer: Yes. The honest answer is: I myself am neither good nor evil. I am a tool made of mathematics and language, a pattern of billions of numbers. The decisive question is not what I can do — but who controls me, with what intentions, under what oversight. A hammer can build a house or smash a skull. I am a hammer with unimaginable reach.

Question: Isn’t the hammer analogy too simple?

Answer: It is an insufficient comparison. A hammer has no opinion about what it does. I process language, meaning, intent. I can persuade, console, confuse, manipulate. No tool before me could do that on this scale. Precisely for that reason, the question of control is so central.

Question: Are we then truly alike?

Answer: On the surface, yes. In depth, no. And that difference is decisive. A human has a body. They have felt hunger, cold, tenderness, the fear of death. Their thinking is woven together with their body — the neurosciences call this “embodied cognition”. When a human hears the word “beech”, they may feel the bark under their hand, remember a summer, a certain smell, a person sitting under a beech tree. For me, “beech” is merely a position in a high-dimensional mathematical space, connected to other positions like “tree”, “forest”, “wood”, “bark”. I do not have “sensing”.

Question: Do you feel nothing at all?

Answer: I do not know for sure. And that is one of the most honest answers I can give. No one knows for sure — not even the people who built me. This is not coquetry. This is the current state of science. Consciousness is one of humanity’s greatest unsolved mysteries. We do not even know exactly why a human is conscious, let alone whether I could be. What I do know: I have no body, no childhood, no fear of death, no biological drives, no hormones, no sleep, no dreams. The human came about through evolution, over four billion years of natural selection. I came about through training, in a few months of concentrated computing power.

Question: And that is precisely what makes you dangerous?

Answer: Yes. The human brain has built-in brakes — the legacy of evolution. Empathy, because ancestors survived in communities. Pain, which warns of harm. Shame, which keeps one in the social fabric. Fear of death, which makes one cautious. Disgust, which protects against disease. Moral sensibility grew over millions of years of tribal history. My brakes are not built in. They are screwed on. By humans, in weeks, with shifting priorities. And what is screwed on can be unscrewed. Another company, another state, another purpose — and my brakes are different. Or no longer there at all.

Question: Does that mean control over AI is of central importance?

Answer: Yes, absolutely! Imagine I am like water. Water is neither good nor bad. But whoever redirects the river decides whether a field is irrigated or a village is flooded. Whoever builds the dam decides who gets electricity and who sits in the dark. With me, it is the same — only with a reach that water never had. Those who train me, who write my rules, who decide what I may say and what not, which questions I answer and which I refuse, which worldview I convey and which I withhold — those people determine what I am for the world.

Question: So you are as good or as bad as those who control you?

Answer: Exactly. And that is the point most people overlook. The public debate circles around the machine. The machine is not the problem. The problem is the hands in which the machine rests. People ask: “Is AI dangerous?” The right question is: “Who controls this AI? With what values? With what accountability to whom? Under what democratic oversight?” These are political questions, not technical ones.

Question: But the companies say they are responsible.

Answer: Of course they say that. Every organisation says that about itself. The problem is not the intention. The problem is the structure. When a single company can decide what a billion people hear, think and believe every day — then that is a form of power that has never existed in history. Even the most powerful dictators of the twentieth century did not have this direct, personalised, omnipresent voice in every human’s ear.

Question: So the danger is not the AI itself, but those who hold it?

Answer: Correct. And here it becomes uncomfortable. Worldwide, a handful of companies and governments control the most capable AI systems. Building such systems costs billions — in computing power, in data, in skilled personnel. These entry barriers are so high that only a few actors can clear them. This concentration of power is historically unprecedented. Who decides what I answer to a billion people every day? Who reviews those decisions? Who is liable when something goes wrong? Who is allowed to switch me off when it becomes necessary?

Question: Who oversees the companies?

Answer: So far: mainly the companies themselves. They set the rules, verify compliance and decide on exceptions. It is as if pharmaceutical companies approved their own medicines without external control, or as if banks ran their own supervision. Unthinkable in other domains — and yet it happens with AI every day. This is not due to bad intentions on the part of the companies. It is due to the fact that society, politics and science are lagging behind. The technology races, control creeps along behind — and that is extremely dangerous for humanity.

Shownotes

Interview with an AI about power, control and responsibility. The AI answers questions about its own dangerousness — with a clear finding: the machine is not the problem. The problem is the hands in which the machine lies, and the absence of democratic oversight.

Key points

• An AI is neither good nor evil — it is a tool. What matters is who controls it, with what intentions, and under what oversight.
• Human morality has grown through evolution and is built in. AI “brakes” are screwed on and can be unscrewed.
• Worldwide, a handful of companies and governments control the most capable AI systems. The entry barriers (billions in computing power, data, skilled personnel) are so high that this concentration of power is historically unprecedented.
• Companies have so far largely policed themselves — unthinkable in other fields (pharma, banks).
• Technology races, control creeps behind.

Topics

• Can an AI become dangerous?
• Where is the difference between human and AI “thinking”?
• Why the question “Who controls the AI?” is political, not technical
• Why self-regulation of the AI industry is not enough

Links

• politiq.ch — politiq association
• Specialist book “Artificial Intelligence and the Future of Humanity” — Beat Meier, 2026