Truth Machines: Synthesizing Veracity in AI Language Models
This addresses the problem of AI as arbiters of truth in domains like healthcare and law, but it is incremental as it builds on existing critiques without proposing a new solution.
The paper examines how AI language models like InstructGPT and ChatGPT synthesize and present truth-statements by operationalizing veracity through data, architecture, and social feedback, highlighting the non-trivial challenge of defining truth in AI systems.
As AI technologies are rolled out into healthcare, academia, human resources, law, and a multitude of other domains, they become de-facto arbiters of truth. But truth is highly contested, with many different definitions and approaches. This article discusses the struggle for truth in AI systems and the general responses to date. It then investigates the production of truth in InstructGPT, a large language model, highlighting how data harvesting, model architectures, and social feedback mechanisms weave together disparate understandings of veracity. It conceptualizes this performance as an operationalization of truth, where distinct, often conflicting claims are smoothly synthesized and confidently presented into truth-statements. We argue that these same logics and inconsistencies play out in Instruct's successor, ChatGPT, reiterating truth as a non-trivial problem. We suggest that enriching sociality and thickening "reality" are two promising vectors for enhancing the truth-evaluating capacities of future language models. We conclude, however, by stepping back to consider AI truth-telling as a social practice: what kind of "truth" do we as listeners desire?