< return to HITL thought experiment

As a species, we have always been fascinated with the power of writing. We have postulated the fall of societies due to their lack of recordkeeping; we have mythologized the ancients who authored our prayers, histories, spells, and laws; we have reveled in the magic that comes from bringing an idea from its inception to a form that can be shared with others. Language has always been an infinitely generative system, and there is perhaps no greater demonstration of this fact than the writing you are reading right now. We are on the cusp of a major shift in humanity’s ability to express itself. The first AI language models are trained on a corpus of human writing, and their impact on our society is far-reaching and still unfolding, with implications not only for authorship, but also for translation, education, and debate.

From a writing standpoint, AI language models have already defied all expectations. Programs such as these write articles, novels, and even poetry; chatbots have become more personable than ever before; the internet is awash with fake news sites, conspiracy theories, and amusing blog posts by artificial comedians.

The ability to use computers as an extension of ourselves, with unprecedented power to analyze and compose language, is a reality that is here to stay. As ever more domain-specific language models are developed and incorporated into new interfaces, we may approach a singularity of sorts, an infinitely recursive expansion of the expressive capacity of our language, driven by human creativity on a technological foundation.

Because of this, we are placing this technology in the hands of the public. We believe that authorship and creativity will not suffer, but will rather gain from the fact that people can leverage the semantic power of these systems.

What will you do with this power? What will it mean to be an author when any twelve-year-old can spin a hundred variations on the same fairy tale, and have them all be reasonably well-written?