Artificial Trust is a collection of essays, offering a framework for understanding complex systems. It's an attempt to explain how algorithms can fail in the things we trust.
We rely on complex systems like our health, our society, and the internet. These in turn rely on mechanisms to govern themselves and to evolve. These mechanisms involve feedback and selective pressure, which in turn rely on some mechanisms of recognition and trust.
The Achilles’ heel of a neural network is novelty. When an unfamiliar new input comes along — something that doesn’t appear anywhere in its training set — an AI is trained to reject, not accept it. …read more
Life-like systems give rise to life-like diseases. When AIs make errors, they often reoccur in systematically recognizable and resilient ways. These are not design flaws, because models are evolved, not designed. Every engineered system has an author, but evolved systems disperse accountability over millions of generations. …read more
With biology in mind, the answer is pretty obvious: if you want AIs to be ethical, the only way to be sure is to apply strong selective pressure. The best way to do this is to apply a swift and certain threat of termination to unethical AIs. …read more
Generative AI is not just a tool, but a weapon. It lets you generate strong false signals at scale. The capacity to generate false signals far outstrips any capacity to detect true signals. In fact, the false signals end up polluting training sets for legit recognition. …read more
The idea of government is the idea of imposing control over something. The Greeks used the same word, cybernetics, to describe two closely related ideas. It meant both “steering a ship” and “governing a society.” …read more
Superintelligence is sexy because it’s the romantic fantasy enemy you wish you had.…read more
Should we try to eliminate bias? Perhaps in the systems we build, and in our own lives? When we see bias as the source of injustice, or as a tendency to turn something right into something wrong, well, who wouldn’t want to get rid of that? …read more
Our brains store decisions, like ingredients in a cupboard. We keep these decisions around to inform further decisions. …read more
Artificial trust closely resembles natural trust, as it reduces the computation needed to achieve confidence in a result. But in artificial trust, this confidence is false, because the results are not reliably valid. …read more