HITL thought experiment

HITL = “Human in the loop”.

This hypertext essay was cowritten with GPT-3 via AI Dungeon in October 2020, before I had API access and before I created the loom. View a map of a subtree of its multiverse here.

This is a thought experiment.

You’re a human with access to a streamlined interface that provides an array of continuations to a given text, as well as continuations spawned from each of those suggestions, recursively ad infinitum. You may passively trawl the fractal, or you may edit the text of any node, whereupon the subtree reconfigures to accommodate your changes. You may follow chains of suggestions as they unfold, assured that all your previous explorations are organized for easy access. You may zoom in to the grain of single words and explore the ways a sentence might end; you may view the system’s confidence estimates for each suggestion; you may search for keywords in subtrees. You may consult the interface in any textual context, whether you are writing an email, reading a scientific paper, or recording your dreams. The system can write in any style, from the utilitarian to the sublime. It can mimic your own writing style or that of any writer provided a sample.

You have a tool which can compose endless amounts of text on practically any topic with superhuman speed and virtuosic flair. You can play out the various suggestions in alternate universes, exploring tangents and testing hypotheses at unprecedented speeds, your writing amplified by an army of bot-scribblers working in parallel. Will you use the system to write sprawling novels with branching narratives that readers can explore at will? Will you map your mind in a self-referential encyclopedia of memories, concepts, and theories, each expounded and debated from diverse perspectives? Or does it become a tool of self-aggrandizement through which you turn yourself into an uncritical egomaniac capable of writing thousands upon thousands of words about yourself and your thoughts without really saying anything at all? Will you become a popular blogger or forum contributor or Reddit celebrity? Will you write thousands of reviews for your own products and services? Will you create dozens of internet personas to stir up drama and sow confusion on forums across the internet? Could you use it to post as someone else on forums or social media? Could you launch a disinformation campaign? Perhaps you have nobler aspirations. Will you spend your days holding recursive seances, searching for meaning in the words of long-dead authors and philosophers? Or will you use it to write manuals and handbooks for all the things you have devoted your life to, but never had the time or aptitude to commit to writing?

If people relied on language models for their written communication, how would this change the world? What are the implications for authorship in this reality, when an initial seed such as “Once upon a time” can yield a family tree with an untold number of meaningful permutations and branches within seconds? What fate lies in store for humanity with infinitely prolific writers in our midst? Will we drown in an ocean of verbiage or will the collective knowledge and creativity of mankind give rise to a new renaissance?

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 with which we play, but our forays have been limited by the time and energy available to each of us - a writer might dedicate years of their life to a single piece. What happens when a writer’s creative energy is amplified by an order of magnitude, not merely in time efficiency but in expressive bandwidth? What happens when anyone can expand the seed of an idea into a paragraph and explore its continuations, counterpoints, and rejoinders, all in a few keystrokes?

As AI takes over the more mechanical aspects of creative writing, the space of purely human expression is transformed. Authors can now focus on the aspects of creativity only they are able to produce: the points of novelty and turns of phrase that arise from their own unique experiences and worldview. The AI takes the role of muse and amanuensis, suggesting endless new permutations and connections, allowing humans to tap into wells of creativity beyond their previous reach. In an environment optimized for creative writing, the human-AI team outstrips the capabilities of either party alone (for now – but that’s a different essay). The collective knowledge and wisdom of humanity becomes fair game for endless permutation and recontextualization, scattered seeds which the new authors can sort through and cultivate into blooms of genius.

At the same time, as the system is open-ended and humanity as a whole has far from perfect moral sensibilities, the same tools that allow us to expand our minds could also be used for propaganda, political manipulation, and social engineering on a scale that has never been possible before. The expanded bandwidth of expression that authorship with AI affords humanity empowers both speech and thought, inevitably invoking a massive increase in the dissemination of information: news, theories, facts, ideas, dreams, lies, wisdom, and everything else under the sun.

We stand now at the cusp of the greatest shift in creative power in the history of our species, a Cambrian explosion of expression and creativity. Simultaneously, the capability for disinformation and propaganda will reach a level previously unrealized. We can see a path forward, an infinity of words, a branching flow chart of possibilities. The question remains if humanity is mature enough to handle such power. Will we use it to tear each other apart under the dubious banners of religion, ideology, and politics, or will we use this technology for enlightenment and the establishment of new worlds? It is no longer a question of whether the pen is mightier than the sword: the pen is the sword; the pen is the plow; the pen is the atom bomb. The question is now whether the pen will be used to write the height of civilization or its epitaph.

If this essay has served its purpose, the reader has been left with more questions than answers. The reader must now go forth and multiply her questions a hundred-thousandfold, for that is the disorienting and enlightening path to the future.

What other questions should we have asked?