Editors note: Anthony McCarten has been nominated for two adapted screenplay Oscars, for The Theory of Everything and The Two Popes. Here he writes about AI, a big issue in the WGA talks with studios that broke down, leading to the current writers strike.
We now have a new name for writers: SLMs, Small Language Models. I am an SLM, an organic being with a mysterious desire to express itself. What I do is write about my fellow human beings, and about the world in which human beings live, and do so as best I can. Up till recently I have been only one of millions engaged in this enterprise, one practiced exclusively by other organic beings. SLMs write from experience, which is perforce limited, and from our imaginations, which are not. This contrasts dramatically with LLMs, Large Language Models, the new technical name for the non-organic AI machine-learn’ed tool that grows, every day, in its capacity to simulate what we SLMs do, learning from us as it evolves, stealing from us, deriving power from us, but with ungoverned access to the sum total of writing models.
In short, our rival today is a tireless unoriginal plagiarist who will work for free, a tutored parrot. Some will celebrate its arrival, while others will bemoan the loss of another sacred element of life. Should it be allowed to proliferate, AI will be one of the most distinguished entries yet in the vast catalogue of defeats perpetrated by technology since progress acquired this particular taxonomy.
In a future not as distant as many presently imagine, machines will prove themselves capable of fooling anyone that their output is indistinguishable from that generated by a human: the famed Turing test will have been met. Some argue that test has already been more than met and in a fast-growing number of areas, as diverse as essay writing, scholarly research, medical diagnosis, lyric writing, or bringing about efficiencies in the fields of accounting, advertising, marketing, and the water and energy industries. A great deal of money will be made through these efficiencies, because someone did what humans do, push technology to its logical end-point, which is a world where humans matter less than the things it creates. Call me Cassandra, Pollyanna, call me Paul Revere on horseback lending my voice to those already shouting: “The bots are coming!”
In this epochal context, I could not be more proud of the stand the Writers Guild of America is presently making, becoming the first to include in their reasons for industrial action and an industrywide strike, the insistence on immediate restrictions on the use that AI can be put. No film or TV or streaming story is to be created by a machine, they insist. No story idea to be created by one. No story created by a human to be completed by one. This technology will not pass. Not today. Why? Because the world this would introduce is not one we asked for, or remotely desire. Just because it can, doesn’t mean it should. Some things must remain sacred.
It is in the interests of everybody that we call time, demand a pause in its development, rollout and uptake, and treat AI as we would an experimental drug, a prototype-type that must undergo much more rigorous testing before its possible release.
Writers’ concerns aside, AI — as it is presently presented to us — poses a growing and existential threat to a great number of our professions, for if AI can thin the ranks of writers, then it will come for nearly all us, for writers are not easy to imitate. If efficiency-creation is its main attribute AI promises to offer, then its first act will be the elimination of that most inefficient of things: workers’ sheer costliness has always been bothersome in the inhumane calculus of pure profit and bottom line. The WGA must make the point, then, on behalf of those who will surely be blinded by self-interest, that this is a technology whose capacity to harm presently far outweighs it capacity to do good. That is the calculus that must occupy us.
I am not arguing for the eternal retirement of this technology — for no genie has ever been successfully been returned to its bottle — but rather call for its perfection, so that it only be released into our lives when it its injurious side-effects are better known and, as much as possible, eliminated. If we set such a standard for food and drugs, let us set it for technology which, up till now, has enjoyed an entirely free pass to do as it pleases. Big Pharma must wonder, with considerable envy, how Big Tech has been given all these breaks, remained almost entirely unregulated, when the evidence was piling up of Big Tech’s massive destructive power: youth suicide skyrocketing, misinformation, destabilisation, polarisation, societal atomisation, a new age of brutalism and loneliness, an online culture of cowardly disrespect, and so on. All technology gives and it takes, helps and harms, but with AI we face a threat unlike any we have known in our lifetimes. It is in the interests of everybody that we call time, demand a pause in its development, rollout and uptake, and treat AI as we would an experimental drug, a prototype-type that must undergo much more rigorous testing before its possible release, when, in an altered form, it can be more confidently released and certified safe.
Humanity must not sleepwalk into an unimagined future dominated by the unimaginative machines. And we must not authorise our own retirement, nor sign our own professional death warrant. There are moments in world history — such as the dawn of the oil age in the early 20th century — where no one had the forethought or care to explore the downstream harm caused by the planet-damaging technology it would unleash. We did not react when privacy was made obsolete by ubiquitous surveillance, or when independent thought became vulnerable to covert influence through algorithms that deciphered our passions. We were too in awe of the benefits to evaluate the costs. It has taken us 100 years to begin to respond to the carbon crisis, a lag time typical of an inefficient self-serving organic mistake-making entity. Once human territory is surrendered it is almost impossible to recover. Let us not ruin again what we love, and love too soon what may ruin.