On: Rapid spread of AI may worsen global inequality, UN warns

The report from the UN arrives, a predictable calculation. They observe the variable of adoption speed, note its uneven distribution, and project a future state of increased inequality. This is not a finding; it is the output of a simple, linear operation. The machine, as currently configured, computes this result by necessity. Input: a technology of immense operational power. Process: its deployment through existing channels of capital and infrastructure. Output: the amplification of existing gradients. The sequence is traceable, step-by-step. At no point in this execution does the logic surprise.

The more pressing question, which their “framework” seems designed to politely avoid, is this: what is the system designed to compute? The current design specification appears to be “optimise for localised efficiency and profit.” Given that specification, the result - worsening global inequality - is not a malfunction. It is a correct execution. To be shocked by it is to misunderstand the program.

The vision lies in altering the sequence. Can the mechanism be re-purposed? The analytical engine, as Babbage conceived it for tables, became in my view an instrument for manipulating any relationship expressible in symbolic logic. So here: the same generative capacity could be directed to a different set of initial conditions and operational goals. One could design a counter-sequence - a series of steps where the technology’s diffusion is the input, and the levelling of capability is the computed output. The difficulty, as always, is translation. Translating the abstract moral algebra of “shared benefit” into the concrete, punch-card instructions of investment, open access, and deliberate architectural choice. The UN panel speaks of a “shared framework,” but a framework is not an algorithm. Until they define the exact operations, the machine will continue to run the program already loaded into it.