The illusion of rationality
Engineers believe they are rational because they write code. But code is not reason; it’s execution. Reason is the act of questioning your premises, tracing cause and effect, and being willing to discard your own idea when the data contradicts it. Silicon Valley confuses logic with truth. That's why slogans like "move fast and break things" sound clever until you notice that what's being broken are the conditions for thinking.
The dogma of scale
In physics, scaling laws have limits. Double the energy and your system doesn't necessarily double its performance, it can collapse. Yet in AI, the prevailing belief is that more data and bigger models will automatically produce intelligence. It's a superstition with GPUs. The belief survives not because it's proven, but because it funds the next round. Investors mistake computation for comprehension. What grows is not understanding, but the electricity bill.
The myth of neutrality
Silicon Valley's prophets claim data is objective. They forget that every dataset is a fossil record of human behaviour, prejudices, omissions, power structures, preserved in CSV. When you feed that into a model, you are not purifying knowledge; you are distilling bias. Calling that "neutral" is like claiming your microscope lens doesn't distort light. It does. You just haven't measured how.
The theology of alignment
The idea that human values can be encoded in loss functions is one of modernity's strangest theological projects. It assumes that ethics can be compiled, that society can be simulated, that disagreement can be debugged. Alignment research tries to build moral philosophy without philosophers. It's not evil, just reductionist, an engineer's wish to make the world behave like software. But the world's source code is social, not digital.
The economy of inevitability
Markets in the Valley reward acceleration, not verification. The faster you declare inevitability, the more capital you attract. "AGI is coming," they say, which really means: "Don't regulate us yet." If science progresses by disproof, techno-capitalism progresses by pre-order. The belief in technological destiny keeps valuations high and accountability low. That's not reason, it's eschatology with stock options.
What reason would look like
A reasoned approach to AI would start where the hype ends. It would ask: What causal mechanisms explain the model's behaviour? What harms appear when this system scales? Who bears the cost of our uncertainty? It would treat ethics as governance, not as a GitHub issue. It would publish negative results. It would prize falsifiability over followers.
The real intelligence test
AI doesn't threaten human reason; our reverence for it does. The danger isn't that machines will start thinking. It's that humans will stop. When code becomes creed, science becomes ceremony. The antidote isn't more computation; it's curiosity disciplined by evidence. That's what separates the scientist from the believer, and progress from noise.
