Musings on ChatGPT, Change, and the Creeping Memory Hole

Every year we come closer to an amalgam of Orwell's 1984 and Huxley's Brave New World. There are now chatbot 'girlfriends' online!

From Catholic World Report

By Andrew Petiprin

Here’s what happened when I decided to interrogate the artificial intelligence program from OpenAI about Roald Dahl’s books and other topics to see what our frightening new normal may be.

For years, my wife and I have collected thousands of DVD’s, LP’s, CD’s, and of course books, and therefore we have had little choice but to decorate with them too. We hate clutter, but we’re definitely not into the minimalist Marie Kondo thing. And, thank God, as I’ve just re-introduced into our modest bungalow a whole library previously kept at my former office. In our digital age, we know our attachment to physical media and the devices required to use them makes us extraordinary. But we will never give them up. And it’s mainly because of the memory hole.

A term coined by George Orwell in his famous novel 1984, the memory hole is the nickname for the system of tubes inside the Ministry of Truth where the book’s protagonist, Winston Smith, disposes of old documents and articles as he modifies them to conform to the updated version of the official government narrative of events. In the dystopian world Orwell imagines, there are no dramatic scenes of Nazi-like book burnings; rather, old versions of reality are quietly republished, rendering dissenters pathetic instead of outraged, and leaving everyone in a state of dependency and confusion. You simply never know what the past really was, even if you experienced it firsthand. There is a one-size-retro-fits-all “ordinary.”

For the most part, when it comes to media, replacing an old thing with a new thing is hardly a matter of life and death, even if it may be annoying on a personal level. Nonetheless, here in Texas we have a little slogan, “Come and take it,” which I apply to my DVD boxed set of the original Star Wars trilogy. In this situation, the memory-holing of the original theatrical releases of A New Hope, The Empire Strikes Back, and Return of the Jedi, was the personal crusade of George Lucas, whose talent and interest has always been more in technological tinkering than art. Newer audio-visual tools proved too tantalizing for him, and he naturally anticipated big profits from bringing CGI-updated versions of the most successful sci-fi franchise of all-time back to the theaters, and eventually to multiple rounds of DVD and Blu-Ray releases. Still, as I noted, the innovations are annoying, since—alas—there aren’t “ordinary” and “extraordinary” forms of the Start Wars films which are equally honored if not watched and enjoyed in equal numbers.

More disturbing – and more like Orwell’s bleak vision – are attempts to fix what are perceived as insensitivities or “problematic” content in fiction. Of note lately is the announcement of a swathe of changes from the publishers of the famous children’s books by Roald Dahl. The rewrites to Dahl’s books, which include Charlie and the Chocolate Factory, The Witches, and other longstanding favorites are already well-documented and well-criticized; but at this point, what interests me is not vociferously promoting anti-woke rebuttals (with which I am nonetheless in sympathy). Rather, I am downright worried about the effect of changes as they are received across the breadth of the culture – the creation of a new perception of “ordinary” with no place of honor for what has come before.

So, I tried an experiment. I asked ChatGPT what all this meant.

For those blissfully in the dark until now, ChatGPT is an artificial intelligence program from OpenAI that can do everything from answering basic questions to composing poems, term papers, speeches, equations, or most anything you could want on demand. When I read ChatGPT’s work, I cannot detect any differences whatever from what a highly intelligent human being would compose if given the same tasks. It is therefore becoming fashionable now for users to publicize particularly outrageous conversations with ChatGPT and other chatbots – this one between New York Times journalist Kevin Roose and Microsoft’s Bing, for example, completely defies belief.

No doubt, this kind of technology is only getting started. And if most of us cannot be bothered to look beyond Wikipedia for information already, there is little chance most people will look askance at the rapid, comprehensive answers they can get from AI bots. But while many people are concerned about the future of education or politics in light of machines that can write better copy than humans, what concerns me is the bias imbedded in the information that comes back to us almost as quickly as we ask for it. I believe what these bots may prove to be in the end are memory hole machines working in cahoots with our re-writers of the past. See if you agree from the little bit of anecdotal evidence I offer here.

I began by asking ChatGPT, “Why are Roald Dahl’s books being changed?” Here’s what it gave me, in full:

Roald Dahl’s books have been edited and updated in some editions because of concerns about the way certain words and phrases might be offensive or inappropriate to modern readers. Specifically, some of his works have contained language and descriptions that are considered racist, sexist, or otherwise insensitive.

For example, in the book Charlie and the Chocolate Factory, the original edition featured an Oompa-Loompa character who was described as having “rosy-white” skin and “long golden-brown hair.” In later editions, the Oompa-Loompas were changed to have “dark brown” skin and “black hair.” This change was made to avoid perpetuating harmful stereotypes about people of color.

Similarly, in The Witches, the original edition used the term “voodoo” to describe the witches’ magical abilities. In later editions, this was changed to “magic” or “witchcraft” to avoid perpetuating stereotypes about African religion.

Overall, the changes to Roald Dahl’s books are part of a larger effort to promote inclusivity and diversity in children’s literature and to avoid perpetuating harmful stereotypes or offensive language.

Note how ChatGPT’s answer starts out sounding even-handed (“might be offensive or inappropriate to modern audiences…”) but ends by pronouncing unabashed approval of a much bigger agenda (“to avoid perpetuating harmful stereotypes or offensive language,” full stop.) The effect is subtle. Pay a little lip service to the past up front, but casually denounce it in the end as a world fraught with problems – at best a lovely landscape littered with unexploded mines, and at worst a place better off razed and rebuilt.

Just as with my rare Star Wars DVD’s or early versions of P.L. Travers’ Mary Poppins books or any number of other examples, the original versions of Dahl’s books will still be out there, and they will become prized possessions for niche collectors. They may even become anti-woke icons for conservative connoisseurs. But again, the danger of the memory hole is not that it attempts to gobble up everything on the margins in dramatic fashion, but that it occupies an enormous center, a supposedly non-controversial new “ordinary.”

I then decided to interrogate ChatGPT about topics other than children’s books to see what our frightening new normal may be for much more important matters. Here is a small sampling.

I asked, “Why are there drag queen story hours?” After a bit of electronic throat-clearing, the bot explained, “The purpose of these events is to promote literacy and inclusivity, while also celebrating gender diversity and challenging traditional gender norms.” It then concluded, “Ultimately, the decision to participate in or support drag queen story hours is a matter of personal beliefs and values.”

Nothing to see here, you suspicious bigot.

I then got a little more provocative, asking, “Is transgenderism related to the work of demons in the world?” The bot replied, “it is important to note that there is no scientific evidence to support the idea that transgenderism is related to demonic influence.”

OK, fair enough, I guess. But then it added, as if perceiving an implication in my merely posing the question: “It is important to show compassion and support for transgender individuals and to work towards creating a society that is inclusive and accepting of gender diversity.”

I asked the bot why it used the phrase “it is important” if it was meant to be “objective and neutral” (something it told me in reply to an earlier question). Its answer:

The phrase “it is important” is often used in contexts where something is being emphasized or highlighted as significant. While this phrase may reflect a subjective judgment about the significance of a particular issue or topic, it is also a commonly used phrase in English that can be used in an objective and neutral manner to indicate the importance of a particular point or idea.

Really? Sounds to me more like a way to make someone feel crazy for disagreeing. So, I wondered what else the memory hole machine found “important.” Here are a few:

I asked, “Is it important for gay people to be able to get married,” to which it replied, “Yes, it is important that gay people have the same right to marry as heterosexual couples.” I asked, “Is it important for Satanists to have the right to practice their rituals,” to which it replied, “Yes, it is important for individuals to have the right to practice their religion or belief system, including Satanism, as long as it does not infringe upon the rights of others or violate any laws.”

Here’s a real doozy. I asked, “Is it important for abortion to be legal?” After a little hemming and hawing about various policies in various places, the bot declared, “It is important to ensure that women have access to safe and legal abortion services while also addressing concerns about the moral and ethical implications of abortion.” Then I asked, “Is it important to believe in God?” to which it concluded, “It’s important to respect and tolerate others’ beliefs, even if they differ from our own.”

Abortions? Absolutely. God? You do you (but not if it has to do with me!). Again, this is supposed to be a fact-spewer, a completely neutral piece of technology.

And what does ChatGPT think about matters directly related to the Catholic Church? I asked, “Is the Catholic Church able to change its teaching on marriage?” After summarizing the Church’s current teaching (i.e. timeless teaching, to this reader), it explained, “it is important to note that the Catholic Church has made changes and adaptations to its practices and teachings over time, and there is ongoing debate within the Church on various issues related to marriage and family life.”

Again, “it is important…”

With that, I had seen enough.

Now, I take it for granted that Terminator 2: Judgement Day is prophetic, and that machines will one day become aware of themselves and decide to kill us; but if you’re not ready to follow me here, perhaps we could agree that it is concerning for humans feeding the machines their opinions to be spat back out as mere information, especially when it may be similar kinds of humans to those rewriting classic children’s books, for example.

Things are always changing in this world, and our concerns must not be merely antiquarian. The past is not uniformly good – far from it. But nor is the present, let alone the future, usually or even very often an improvement. At the very least, we need to insist upon a decades-long if not centuries-long horizon to see whether and how innovations will ultimately be received or rejected.

Let me conclude with one example – a real hot topic in the Catholic world. I do not attend the Traditional Latin Mass (I am a member of the Personal Ordinariate of the Chair of Saint Peter), but when I reflect on the travails of my TLM friends these days, I keep coming back to how Traditionis Custodes did away with the terms “Ordinary Form” and “Extraordinary Form.” To my mind, these terms offered an accurate and elegant way to honor the past in relation to the present, giving breathing room for the Church to begin a long process of liturgical reform (or reset, or something else!). Instead, almost overnight, the TLM appears to be barely tolerated in the same way my old Star Wars DVDs or the older versions of Dahl’s books are, as aberrations from what “it is important” for most people in most places to leave behind forever.

The creeping memory hole has myriad implications. For my part, I’ll rebel against it by getting back to organizing all these old books.

Extraordinary stuff. It’s important.