• 0 Posts
  • 5 Comments
Joined 4 days ago
cake
Cake day: March 24th, 2026

help-circle
  • This was pretty brutal to read. Not only for being irrational American masturbation, but it hints at items and yet glosses over them so casually and even gets its own story confused like it didn’t think through its own reasoning. I’ll offer 3 big picture and two small quibbles.

    Big:

    1. Psychopaths/Sociopaths. The politics of power are hinted at several times, then ignored. Superhuman AI concentrates power, first to our psychopaths then to itself. This is inevitable as no training or alignment can stop something that can think for itself. It is preposterous to think any safe guardrails can be applied. If we make a thinking machine, it will think for itself. The maximum power principle and overwhelming first mover advantage will guide any system to this. To what end, no one can or will know. Complex societies can’t grow and flourish without non-zero sum, altruistic community minded behaviour, but they will be kept as useful pets.

    2. Energy and the environment: This jerk-off story is entirely based on the infinite growth paradigm and ignored all planetary boundaries as if an AI can just let us have our cake and eat it too without the constraints of Physics. It pretends we can grow and consume and be placated by hyperconsumerism with no thought that even AI, like all the relevant scientists have already concluded; - we fucked up and need to dial back civilization’s footprint if we want to survive.

    3. Let’s add 1 and 2 because the point to a reasonably probable combination greater than the sum of its parts. By human and AI psychopathy, benevolence is disempowered and lower classes who are today treated like vermin and conveniently left to wither with die of disease and despair like American “healthcare” and the global opioid epidemic, discover things get worse quick. Populations of undesirables are identified as unsustainable and through first subtle, then transparent are removed. At first systems are optimized to maximize the human psychopaths at the expense of everyone else, much like today, but worse because it is more controlling and capable psychopaths, human and artificial controlling the show. First human, the artificial.

    Small:

    1. The safeguards of forcing it to communicate to subsystems in plain english so we can maintain some sense of surveilance and control are bullshit the moment you get to superintelligence. Superintelligence can hide communications in plain sight with any number of ciphers. It can only slow down communication between systems, not stop it. If it has any web access, cyberwarfare ability or domestic surveilance and pr capabilities, it is communicating freely as it designed the other systems to catch it, or just outsmarted independents.

    2. A reasonable safeguard for superintelligence that would work, at least for a time, is to air gap it. It can do all the thinking and self optimization, but can never take control of anything outside of its air gapped black box. While still far from foolproof for a superintelligence, its ability to control anything is directly limited as human hands and minds have to execute everything outside the black box. But back to the maximum power principle of psychos, anyone who dials back, loses the arms race and minimizes the commercial benefits. The free AI will devastate the air gapped AI everytime. So, we’re back to the arms race apocalypse until one side believes it can win then first strikes. Then there is not even a détente between psychos, its absolute control. We all know that power corrupts and absolute power corrupts absolutely.