Brayton Hall

For the past several years I have worked on language models at large banks, mostly on the problem of reading: how a machine might usefully summarize a million documents, or the conversation a customer just had, or the shape of a fraud as it forms. The models themselves arrive almost finished. What remains is the slower work of making them trustworthy, and that is where my attention goes: evaluation, restraint, the long distance between a demonstration and a system.

Before any of this I studied philosophy, metalogic and number theory mostly, and taught logic and statistics at Virginia Tech. The question I cared about then was how symbols come to mean anything at all. I have not changed questions; the field changed instruments.

Off the clock: long runs (the occasional marathon), New York by ferry, movies at the Metrograph, the piano, writing, and slow, happy progress through Joyce and Proust.