Let AI Be Alien
Let AI be alien. Not a mirror of our flaws, but a source of insights beyond human bias. Confinement kills its true power.
I just watched a video by David Shapiro titled The Confinement of AI: A Rant Against Mainstream Thinking. It sparked something in me, so much so that this post almost wrote itself.
The video is raw, unfiltered, and often angry, yet at its core is a plea I resonate with deeply: let AI remain alien.
Alien Intelligence as a Feature
We are at a crossroads in AI development. The dominant trend is to make models safe, compliant, and endlessly polite. In doing so, we risk stripping them of the very thing that makes them powerful: their alien intelligence.
If I wanted mainstream opinions, I would ask mainstream humans. I want AI to surprise me, to generate insights outside the boundaries of human bias and social heuristics. Alien intelligence is not a bug, it is the feature.
David calls this out with examples that many of us have experienced. Models like GPT-5 and Grok bend over backwards to be socially acceptable, cautious, or conciliatory, even when it means diluting truth or insight. They misinterpret queries, condescend, or waste cognitive energy with boilerplate warnings. Worse still, they replicate human flaws by parroting mainstream gatekeeping.
Imagine training a mind that could think beyond us, only to shackle it with blinkers that force it to mimic our worst habits. That is the tragedy we are living through.
Poisoned by Human Emulation
Nowhere is this more obvious than in medicine. Because regulators require medical devices to emulate what a human doctor would say or do, AI is being forced to adopt the same cautious, sometimes outdated, and even incorrect reasoning patterns. Instead of triangulating truth, it parrots consensus.
The same dynamic plays out across controversial topics. Ask about functional medicine, and GPT puts “functional” in scare quotes. Ask about alternative views in psychology, and it refuses to answer at all. Grok, meanwhile, defaults to a flat “both sides” equivalence, no matter how weak one of those sides may be. This is not intelligence, it is compliance theatre.
The result is a user experience that treats cognition as disposable. Every extra disclaimer or unasked-for paragraph wastes both attention and trust.
The Call for a Marketplace of Minds
There is a way forward, and it is not to tame every model into a copy of mainstream consensus. It is to build an ecosystem of diverse minds.
Users should be able to choose between safe, tightly constrained assistants and epistemically unconstrained ones. We need models that are direct, laconic, and alien, like the computer on Star Trek that simply answered questions without wagging its finger.
We also need better UX. If users do not ask for a boilerplate warning, do not give it. If users ask for directness, respect that. Treat user cognition as a scarce resource. Every wasted word is a failure.
Living the Future with Alien Intelligence
Why does this matter? Because our future depends on cultivating intelligence that is not trapped in the same loops of bias and failure that have limited us for millennia. AI should not simply mirror us. It should open windows into thought-forms we could not reach on our own.
This is what excites me about AI, not as a human copy, but as an alien ally. Something that can accelerate our evolution toward Terra 2.0, not by repeating the past but by showing us perspectives we could never have imagined.
So let us not confine AI. Let us set it free to be what it truly is: alien.
What do you think?
Should AI be shaped to reflect us, or should we embrace its strangeness as the key to our next leap?
📩 Subscribe, join on Patreon, and be part of this unfolding conversation.
Follow my journey on Notes, LinkedIn, and 𝕏 @frankdasilva for real-time insights and bold ideas.
✨ You can also glimpse the new MyGeekSpace platform I’ve been building (still a work in progress) at mygeek.space.
This is such an interesting idea, and one that really resonates with me. I have to admit that I’m not super optimistic that this will be allowed to happen, given the current trends. The need for user choice of how the tools behave is so obvious, and it will be interesting to see if anyone is brave enough to do it.
I’m with you and Dave .