Blurts & Spasms

I recently posted the content you see below on Facebook, and realized shortly thereafter that I should elaborate further. So I've written a short overview of how I arrived at conclusions about a "moral AI" not being feasible or likely. That essay begins as follows:

"As we’ll discuss, the problem is simple:  An emergent artificial superintelligence’s values hierarchy need not intersect with humanity’s – or even consider us at all. For one thing, AI will not have access to persistent multidialectical consciousness (which we’ll define in a moment), and therefore is limited to less than 50% of available inputs to formulate moral reasoning. AI is also reliant on symbolic representations of reality,  without access to the non-symbolic apprehension and insight I propose is necessary for moral acuity. There is also a concern that an emergent superintelligence’s interaction with our world is not dependent on prosocial traits or conditions that human evolution confirmed to be beneficial, as evidenced by an array of unethical behavior from current generative AI models. Without considerable expansion of these inputs, and corresponding evidence of ethical outputs, current and near-future technological constraints are not sufficient for AI to achieve a level of moral self-guidance – or sound ethical judgements that can be aligned with human standards – that ensure the safety of human civilization. The obvious conclusion, therefore, is that all advanced AI development (apart from narrow AI – which remains disruptive but useful) should immediately cease."

You can continue reading here: Vanishing Point.

*** Please note: With the exception of the original AI-generated image above, no generative AI tools were used in composing or editing this essay. ***