- HubPages»
- Entertainment and Media»
- Movies & Movie Reviews»
- Science Fiction & Fantasy Films
Crunching the Numbers of "Sunspring" (2016)
Simple Ways of Assessing What's Important
Keep in mind that at the time Benjamin wrote "Sunspring," Benjamin was an LSTM (long short-term memory) recurrent neural network. Most who have reviewed the film have admitted that as a film, it is a challenging text. Amanda Kooser, reviewing "Sunspring" in 2016 for c|net, said it "is exactly as weird as you would expect."
But as a text, what might this screenplay say about humanity? And specifically, what might it say about humanity as a lens on the texts Benjamin "read" in preparation to write "Sunspring?"
The Proof of the Product
The wonderful and disastrous thing about these neural network writers is that what goes in comes out, sometimes in unmediated ways.
Consider the case of Microsoft's Tay, a chatbot that quickly became racist and which Microsoft subsequently took offline.
The Basics
While Tay developed badly, what Tay demonstrates is, when unmediated, these neural networks can reflect in raw ways their human-generated input, not unlike human children.
It is crucial to keep in mind that Benjamin was not subject to public conversations. Instead, Benjamin was fed sci-fi screenplays. Sci-fi screenplays are a specific lens on humanity, and that narrow view impacted what Benjamin generated. But "Sunspring" is somewhat exciting and reflective, arguably, of the broader human condition.
So, who appears in a scene?
Within "Sunspring" there are three characters. Not surprisingly, We can read the three characters as being involved in a love triangle. Considering the scripts Benjamin read came from the 1980s and 1990s, you can probably guess that this particular triangle is a hetero-normative one with two males and a female.
At this level, Benjamin seems to be reproducing the status quo surprisingly well for a neural network.
Technical note: For those who appreciate the in-depth. Consider that this screenplay is a collaboration between Benjamin and Ross Goodwin. Goodwin mad changes including converting character names to single letters in the screenplay, in part because Benjamin did not handle proper names very well, according to Ken Miyamoto in an article for Screencraft.
The Product as Our Proof
We know the characters.
But what do they have to say? What could Benjamin possibly put in the mouths of humans?
Once Within a, Twice Within a, Thrice Within a
In the style of Benjamin, I have decided to look at the dialogue of "Sunspring" using statistical analysis. Statistical analysis is also the base of a lot of natural language analysis and fuels a lot of machine learning in the natural language field.
Some notable results emerge from the analysis of "Sunspring" at the level of simple counting. Here again, we have to keep in mind the parameters of the texts Benjamin read, sci-fi. Science focuses on knowledge, and knowledge plays a significant role in sci-fi as well.
It is never surprising to find a lot of article and pronoun clutter at the top of language counts, so I have omitted those. Despite those omissions, I have preserved the ordering in the table of the top ten words.
"I" is the most common word in "Sunspring." We can read that as significant. Relationship dynamics generate a lot of energy while watching the film, but ultimately the film's aboutness can be read as centered on the isolation of the self. Elements resulting from the non-poly triangle and the presence of monologues in the film reinforce this isolation reading.
Word #10, "know,' in the table opens up another significant expression within the film. While we might connect this to the sci-fi genre, it is part of the more prominent phrases within "Sunspring," some expressing complete grammatical sentences.
The top phrase containing three words is, "I don't know." And the top phrase containing seven words is, "I don't know what you're talking about." These high rankings resonate at so many levels for this film. Consider the actors attempting to perform words generated by a neural network. Consider a neural network reading texts generated by humans to generate a text to be acted out by humans. Consider yourself, a human, attempting to understand a film performed by actors from a screenplay generated by Benjamin after reading texts generated by humans. But perhaps most interestingly consider ourselves, all human, attempting to understand the other people in our lives: our boss, our lover, our friend.
Benjamin could have written any number of things. But the "Sunspring" continually echoes with the solitary self-expressing their awareness of their inability to access the meanings and expressions of others.
At nine minutes, "Sunspring" is a fascinating first piece from years ago.
Technical note: Lack of understanding is the pivot-point of many assessments of machine learning concerning natural language. This consideration does not begin to address what "understanding" is and if it is possible. The inability to know what language users are referring to is the thesis of W.V. Quine's Indeterminacy of Translation, which arguably, holds for same-language users.
Statistical Analysis of Words in "Sunspring"
Order
| Word
| Occurences
|
---|---|---|
1
| I
| 45
|
5
| he
| 25
|
7
| you
| 20
|
9
| was
| 17
|
10
| know
| 14
|
Statistical Analysis of 3 Word Phrases in "Sunspring"
Order
| Phrase
| Occurences
|
---|---|---|
1
| I don't know
| 10
|
2
| to be a
| 5
|
Statistical Analysis of 7 Word Phrases in "Sunspring"
Order
| Phrase
| Occurences
|
---|---|---|
1
| I don't know what you're talking about
| 4
|
General Likability
Google Users
| IMDb
|
---|---|
2.85 / 5 (calculated from 5.7 / 10)
| |
86% like
|
© 2019 EC Wells