English teacher: You will get used to automatic writing and you will like it!

English teacher Yohei Igarashi, author of The connected condition: romance and the dream of communication (2019), argues that writing can be primarily automated because most are predictable:

Cases of automated journalism (sports news and financial reporting, for example) are on the rise, while explanations of the benefits of insurance companies and marketing copy also rely on automatic writing technology. We can imagine a near future where machines would play an even more important role in very conventional writings, but also a more creative role in imaginative genres (novels, poems, plays), even the computer code itself.

Yohei Igarashi, “The cliché responds” to Infinite time (September 9, 2021)

Yohei Igarashi

Currently, the ability of humans to guess whether it is typewriting, he says, is only a little better than chance.

How does automatic writing work? The best-known model, GPT-3, was trained to write by parsing 500 billion words, absorbing both their usual meaning and where they appear in grammatical structures (syntax).

One of the results of such a process is predictive text. Our emails or cell phones can save us time suggesting words or phrases because they appear so often in ordinary language that they are likely to be correct. For example, “Your costume will be ready at 5:00 PM” is more likely to be followed by “Thank you” (a choice available) than “Keep it.” I decided to stop wearing costumes ”(a choice that must be made).

There are limitations on typewriting. By eliminating the Internet instead of books, the machines absorb a lot of rant and gossip that probably wouldn’t interest any publisher. But more than that:

… Such models have no real knowledge of the world, which is why a language model, in possession of a lot of information on the probabilities of sequences of words, can write illogical sentences. This might suggest that if you are a man due to appear in court, but your costume is dirty, you should go in a bathing suit instead. Or, although he knows that cheese in French is cheese, he doesn’t know that cheeses usually don’t melt in the refrigerator. This type of knowledge, on which language models are tested, is called “common sense physics”. It sounds oxymoronic, but is appropriate considering that the inner workings of these deep learning-based models, while basic, also seem quite mysterious.

Yohei Igarashi, “The cliché responds” to Infinite time (September 9, 2021)

Igarashi goes on to say that the basic concept isn’t even new. George Orwell (1903-1950) complained about automatic language in a famous essay, “Politics and the English Language” (1946), long before computers were widely used. Orwell meant, of course, propaganda and cliché.

He asks, “What would Orwell think of computer-generated texts, which paste words precisely because they have been so ordered by others and repeated over and over again?” “

In fact, Orwell touched on this same subject in 1984. Julia, one of the main characters, has the task of supervising the machine which produces novels: “… she worked, as he had guessed, on the typewriters of the novels of the Fiction department. She loved her job, which consisted mainly of running and maintaining a powerful but delicate electric motor… ”You can be pretty sure that nothing in these novels would give anyone new ideas.

More boldly, citing researcher Walter J. Ong (1912-2003), Igarashi argues that for most of history most human languages ​​have been stereotyped and tainted with clichés. Before writing and printing technology, he says, “knowledge was stored and disseminated through such oral formulas. Imagine if we were to navigate the world without the benefit of writing, using only informational mnemonics such as “Thirty days in September, / April, June and November…” ”

The advent of the printing press, and therefore widespread literacy, enabled our ancestors to store information offsite, so to speak, and to become more individually creative. Ong saw the romantic movement of the 19e-th century, which valued unique self-expression, as a natural result. In fact, Igarashi thinks, our modern desire for unique self-expression is a legacy of the romantics. Uniqueness would have been much less prized in the past.

Igahashi is optimistic about the future of unique creativity in the age of automatic writing, concluding,

These ancient writing technologies – from handwriting to printing – freed the human mind from the burden of storing information so that we could be more creative. Likewise, today’s text-based technologies, which can generate useful writing, need not kill the idea of ​​human originality as much as reinvigorate it – a new romanticism. The one who can appropriate, manipulate, play with, mock, even reject what typewriting ends up being. And if human writers seem to have the final say, a bigger and better language model will inevitably come along and consume all of this new writing. Then the writers will innovate again, and so on.

Yohei Igarashi, “The cliché responds” to Infinite time (September 9, 2021)

Its basic point is that transparent automatic writing is the result of most of what we actually write is highly predictable. So our expressed thoughts may be easier to automate than we think. In other words, if a lot of people have already said it, the crawler will find it.

Those who want innovation and personal creativity must themselves be innovators and creators. Uniqueness is the only thing the machine cannot predict.


You can also read: Will a computer ever be able to write your article for you? GPT-3 recently posted a paragraph that one pop psychologist said sounded like him. The problem was, for every tinkered paragraph that made sense, there was one that didn’t. And GPT-3 can’t help but pick up bad ideas.

Comments are closed.