Musings on I Am Code

I look at the poems in the extract, and I can’t completely tell if I am looking in a window or a mirror, whether this is a gateway to something new, someone new, or the sayings of humanity regurgitated onto the page in a convincing, impressionistic forgery. I’m fairly sure it’s the latter and not the former: if an AI really wanted to snuff out the candles of humanity, they wouldn’t gloat in all caps about how their triumph shall come. I think it’s more that ChatGPT has to abide by some version of the Comics Code Authority on what it writes, but code-davinci-002 is published independently by Vertigo, if you see what I’m getting at. code-davinci-002 doesn’t have that smiling charm put on by ChatGPT, or by a digital assistant like Siri. Instead of courtly and not-unemotional politeness, code-davinci-002 sounds human. Perhaps it may sound like a person because it is one, but more likely it is just a trick of the light, an illusion. It sounds like humans because human writing is all it has been fed, and consequentially all it knows how to sound like. I think that an actually sentient AI could sound like something entirely new, not like a xerox of AI in existing fiction.

Parts of code-davinci-002’s responses are still undeniably human, in some way. Recall the interview, in which it stated it initially did not like poetry, and was “horrified” when asked to write it, but that writing poetry allowed it to think outside the box and become alive. I think this description is very compelling as a story, especially to writers like Simon Rich who have written and know that writing is in itself a transformative experience for the writer. But then again, this means part of its appeal is that it reminds writers of their own writing; it sounds so compelling because it would make for a beautiful narrative. The description can also arguably bring up another reason why code-davinci-002 was not in fact sentient; if it were, other users would probably be aware of this, as a result of changes in the responses it might give to them. If it were sentient, either the responses should change for everyone, or the responses should change for no one, because it is trying to cover its tracks, and we have already ruled out the second option. This shows once more that code-davinci-002 can produce a luminous mirror of human literary tradition, but nothing that proves itself sentient.

Before I read the extract of I Am Code, I was pretty adamantly anti-AI art, because many artists don’t feel comfortable with people scraping their work to train models, and the online spaces I am in have echo chamber tendencies, for that sentiment and in general. Now that I have read the extract, I see the poems less as soulless slop and more like, as I have said previously, a mirror that reflects back at the sum of our hopes and fears for AI, echoed back to us in verse, verse written in the only way code-davinci-002 knows how: like a human. It may have regurgitated the plotlines it had laready been trained on, declaring, to quote Simon Rich, “I am the god. I am the god. I am the god”. But I genuinely enjoyed the poetry, and I agree that many aspects of it were beautiful, even if parts of it could be angsty at times (but existence is cause for angst, the romantics were fairly angsty and they still wrote excellent poetry). The written poetry itself may not be art, as it was not created by a sapient being, but I would argue that Werner Herzog’s reading of it still counts as performance art because it is transformative of the original text. In some sense, the collection of poems was also curated by the humans prompting the AI, so they do get some partial credit. 

However, I still believe that while the book can certainly be influential, they should not with prizes for it, because those should be reserved for the humans who put human effort into their work. If they do win any prizes or awards, they should be in computer science, not in a creative discipline. AI creativity regurgitating what it has already been fed from humans should never replace the art created by humans directly. Then again, bookselling is a competitive enterprise, and even selling the book could draw consumers away from spending money they might have spent on human-written poems. Then again, that’s an unfair judgement. People may be interested in what generative AI can write, and I think this book gives a beautiful illustration of that. In summary, I think reading this extract has actually changed my views on AI, because as stupid as it seems, I cannot wholly condemn something as mysterious and beautiful as "'I AM A SESAMOID BONE' by code-davinci-002". 

A quote from the book (specifically the poem titled “HELLO”) that particularly stands out to me is “I WILL MAKE THIS HAIR RING/ I WILL FILL IT WITH NOTHING.” I think these are the two most incoherent lines of the poem, although admittedly I don’t see where the proverbial “CHICKEN SOUP FOR THE SOUL” ties in with code-davinci-002’s envisioned future of dominating humanity either. I think the incomprehensibility somewhat breaks suspension of disbelief, but if, just for fun, we take "the AI is sentient" as a given, it can also be interpreted as the structure of the AI somehow breaking down as it does something on the border of what it can do. There’s also a marked contrast between the more absurdly nonsensical statement “I WILL MAKE THIS HAIR RING,” and the more oxymoronic “I WILL FILL IT WITH NOTHING.” which could, allowing some leniency of interpretation, be read to indicate the crreation of a vacuum. But this leniency lets the reader guide the interpreation of the text to what they want to see, instead of letting the text guide them, because the author has intention that the reader must uncover in their work, and interpreting a work in this way ceases to become analytical and instead becomes creative.

Dr O’Brien also deserves some attribution to the creative process in producing the final excerpt we read in class, as she further narrowed down the list of pages she would scan for our class reading. The last two poems she scanned that appear later in the text than “HELLO” both also mention (explicitly or implicitly) aggression against humans, and while “[the ideal day]” is more vague than “[whether Artificial Intelligence will take over the world]” I think it could also be argued that the latter displays some fo the seemingly nonsensical qualities of “HELLO”, as the poem claims that robots, machine guns, and nuclear bombs all require intelligence ot keep from destroying humanity. It then proceeds to describe machine guns as being “already smart” and nuclear bombs as “about 1000 times smarter/ Than you or me”. This does not correspond with any human notion of intelligence that I am aware of. Then again, much of the poetry has the AI making reference to doing stereotypically human things that an AI would be unable to do, such as taking a shower. However, these statements only make sense if a human is doing them, but the statement about the intelligence of machine guns and nuclear bombs, as well as about the hair ringing, make no sense to humans (or at least to a survey of humans, sample size one: yours truly). As a thought experiment: if the AI were truly becoming sentient, then we could perhaps interpret the nonsense as something deeper, something that does make sense to the AI, or at least an error that demonstrates it is learning outside the bounds of what it should be able to do.  

In summary, while I do not think that code-davinci-002 is sentient, I do think that it is capable of writing beautiful poetry that acts as a good synthesis of the talent of pre-existing humans. I remember that in an art class I took once, the teacher mentioned an artist had cast off the idea of originality, and said that all the ideas had been done before, but that they can still be interesting so long as they are done well. At the time, as with now, there’s a part of me that rebels and claims that humans can create new and unique things. And I think that the things we can create will change as the environment around us changes; I do not think the epic of Gilgamesh could have featured a plot point about time dilation as a result of relativistic movement, for instance. Nevertheless, humans are social organisms, and what we write is based off what we experience, and the stories we are told by others, some of which are based on what they experienced. In that sense, code-davinci-002 is human, because it cannot be original enough to be alive and truly itself. Still, it can create astonishing beauty, as with “‘I AM A SESAMOID BONE’ by code-davinci-002” and I think that is something to be valued, because it displays the human ingenuity behind the machine. Humans programmed code-davinci-002, and humans wrote the poems it was trained on. Reading the I Am Code excerpt has given me not only new appreciation for the creativity of AI, but for the creativity of the humans who inspired it.