Using AI for Writers, A Computer Scientist's Perspective
Non-fiction - by Steven Winston
Graduation season is in full swing. Young people are eager to join the adult world and start to take up the mantle of dreamers, doers, and inventors of things. Each has their own destiny, but they all listen to the same tired speeches about how they are the future and the hope of the past.
But what does that future even mean? Can you honestly name a time in history when humans have successfully stopped the future from happening? It is inevitable, as my old computer science professor used to say. Technology has no moral compass, no good or bad, nor a hope of stopping, nor slowing down, nor does the cadence even pause to ask how it can be used. Technology is as plain a concept as any law in the sciences or mathematics. Invention is discovery, application is what people do with the inventions.
Yet when a new generation of workers, trained on how to build and create tools that far exceed any in existence, enters a workforce that uses the worn-out tools of the past, the new generation must compete with their peers for a space at the table of relevance. Their new tools might be envisioned as a nightmare to some, eradicating whole industries on their way to acceptance. Other tools might enable entire industries to perform better.
What is the promise of the next generation, or the one after it, into infinitum? Science Fiction authors have envisioned their future world enough times, yet how do we get there? Most importantly, how can the experts of an industry allow and guide their industry to grow for the next generation if they fight against any new tool or technique? Where is the acceptance necessary to attempt to create the world which we love to read about in our Science Fiction books?
Allow me to provide an extreme: In the movie "Finding Forrester," Sean Connery implores a bright student to throw away his typewriter and pick up a pencil. Write once without editing, then go back and edit on the typewriter. I secretly thought that trick might even help my own writing. Sadly, my skill set only ever extended to a keyboard and the logical analysis of programming, on which I've built my life.
But think about that for a moment. The ideal of that fictional author was to throw away technology and pick up a pencil. Do you think that author learned that trick on quill and parchment as his mentors eschewed the pencil for it being too quick? Go back a few generations of authors, and maybe we'd find authors pushing away the printing press with familiar arguments of the contraption taking their jobs or not being the "correct" way to write.
Writing has a long history. Those who claim it as a profession have long left the caves of their ancestors. In those caves, humans told their stories, their news, their knowledge, and their hopes and dreams. We as a society have benefited from the descendants of these storytellers, teaching us everything from how Sisyphus pushes a rock up a hill to Asimov asking the final question. Each generation gives us a thread to hold onto and think about the future. The embrace of the latest technology comes from messy adaptation and a lot of scared assumptions.
So as you think back to your own graduation and the path you took when you heard the words of someone saying "you are the future and the hope of the past," remember that the future is not something to be feared. It is something to be embraced and pushed forward by those able to inject the needed knowledge, wisdom and moral fortitude necessary to adapt new technologies and techniques. Remember that the tools they use to create that future are not limited to those that came before them. The future is unwritten, and the possibilities are endless.
Artificial Intelligence has received a lot of press as of late. As one of the more prolific writers, Asimov has a story to cover many directions that AI might take. He even imagined one called "The Last Question" in which someone simply asks, "how do we reverse entropy?" The death of the universe when it finally fell to pure entropy was met with that lone question finally having an answer from the computer that was crunching away for all history. "Let there be light."
As masterful as that story might be; it captures a few things right and a few things not quite right about what AI is and how it works. AI is a quest to mimic how brains work. It is simply asking what is learning and how does learning work, then mimics that process. So a critique of Asimov's might point out that his AI would only learn from the inputs it has available just as well any biological brain is able to do. Thus, it could never get information about how entropy can be reversed via the only insular thinking it can do. Experimentation, or at the least, gaining new knowledge is missing in the story that would give that AI a chance to figure out how to end up with a biblical ability of uttering 4 words.
Without getting into the details too deeply, AI in general is going to be great at a few things that are generally well understood and defined. If you can conceive that there's enough of a body of knowledge to instruct a class of students in a subject, then AI is probably going to be really good at it. In the industry, we talk about measuring AI performance against human experts and then, to surpass human ability, gathering a group of human experts and allow them to come up with answers. When the AI progresses to answering questions better than the group of human experts it gets really hard to make advances. After all, we're human, who is left to teach the AI new things in that subject area? Sure, we pit one AI against another and they tell us they are better; but in truth, they are beyond our ability so far that in some endeavors we hopelessly wouldn't be able to tell the difference.
Let's talk about an easily grokable discipline. When Gary Kasprov, at the time the world chess champion, faced IBM's deep blue, the world thought that no AI could compete with humans. The genius of Kasprov resigned his first match in '96; he went on to win 3 matches and drawing two leaving the official score as 4-2. Humans rested uneasily at how closely we came to losing. '97 there was a rematch; and ever since move 8 in game 6 when the machine sacrificed a knight and left Kasprov defeated a mere 11 moves later did the final straw break. Humans have never been better than computers again. Now, we have a new generation of chess players brought up in a world where an AI can teach humans how to play chess better. Whole strategies are envisioned around and by computers. Our most recent world champion Magnus Carlson withdrew this past year from competition because the highest levels of human chess are governed by who has memorized the lessons and theories of the AI. The tools have become very powerful and have brought chess to the world with an easier and more astute teacher. Hikaru Nakamora, a super grandmaster, recently stated that the students brought up on AI are better than the best of the past. There's even a quantifiable ELO score system used to rank people and machines which also shows proof that chess has progressed ever since machines took over after Kasprov's defeat. The human players are better than they were before the AI existed. Chess is also more popular than it was before the AI and more approachable as new tools are available. One doesn't need to find a titled player to teach you how to play the game; instead learn from the carefully created instruction machines as they refine your game and bring you the ability to enjoy it on whole new levels.
What then can this teach us about how AI might work in other fields? Maybe if we embrace it, it too can make us better in our own disciplines. Maybe if we embrace the paths of the next generation in providing us new tools, we can guide them from wisdom and experience rather than inhibit them with scorn and derision. Maybe, the level of the human written word would improve, providing more access to those less skilled with pencil and typewriter if we don't punish any effort that doesn't follow the dogma of the past?
Embracing technology provides a sacred chance to shepherd in new technologies and techniques. Instead of questioning what to do when technology takes jobs, question instead how you plus the technology can make an end product better.
Maybe you need an AI to create a first draft and then use your expertise to edit only, by which strategy you will craft original stories that the AI couldn't do on its own. Maybe, as I think most likely, you will find help in teaching an AI what your individual voice is by letting it learn on your corpus of work and enabling the AI to cure writer's block in the blink of an eye writing the next sentence in your voice as you might have done yourself. You can then concentrate on living in your fantasy world as you type. Free from the burden of forcing various thought trains to flow into one another and make sense of each other; later to draft the sequence together as expertly as one might weave a cloth.
So in the spirit of the graduation season, allow yourself to dream of what possibilities might exist for writers if you had control over the future of the AI direction. How would you, oh experts of your industry, want to use the tools of the next generation to enable them to join you in becoming experts, to enable them to learn to write better? This is an exciting time. It's time we learn what is possible and what we can do now with the tools of the new discoveries. It's time for the world of writers to graduate.