Have you played this game?You can rate this game, record that you've played it, or put it on your wish list after you log in.
Playlists and Wishlists
RSS FeedsNew member reviews
Updates to downloadable files
All updates to this page
About the Story
It began when my mother's goldfish died on the dining-room table. I was about to tell her how I lived; my girlfriend and I were about to see the Wauwatosa Thunderdivers. We were dancing on the beach; we were singing songs no one knew they had. An experimental story about a lifelong romantic relationship.
97th Place - 26th Annual Interactive Fiction Competition (2020)
|Average Rating: |
Number of Reviews: 9
Write a review
So YWTMaFAYTaW – I feel like I need an acronym for the acronym, why don’t we just go with Thank a Werewolf – so Thank a Werewolf’s About text says that this is a “collaboration” with the GPT-2 text generation algorithm, which involved feeding the program bits of the author’s (static, I assume) fiction, then curating and arranging the output to form the game. I went into the experience deeply skeptical that I’d find anything interesting in it, and I have to confess that nothing in the game wound up changing my mind.
I can’t really talk about the story or the premise since this is all algorithm-driven gobbledygook – the blurb’s claim that it’s about “a lifelong romantic relationship” I think must be a joke about the relationship between the author and their work, rather than about anything depicted in the text. Format-wise, you get a couple paragraphs of prose then one or two hyperlinked lines at the end that lead to the next bit of the story; I assume there’s branching but given the lack of narrative coherence, much less cause-and-effect, I’m not sure what difference or impact this would have. Most passages have one or two footnotes, which expand when clicked: I think these might be the prompts used to generate the text, since they often used the same words as showed up in the main text, but were uniformly more coherent and interesting than what was above.
If there’s not a consistent narrative, there’s perhaps a somewhat more consistent tone, beyond of course the omnipresent surrealism – most of the generated text feels like it’s about younger people, late teenagers to twentysomethings, sad about stuff and unsure about what they’re doing (I mean, fair enough). And there were a few scattered bits that had some zing to them. This was probably my favorite:
"I began to write the descriptions of my wishes. Letters with descriptions of things I wanted to happen, wishes that were already in the cards. An ambulance for a sick child. A dog for its owner. A car so nice it needs a full-sized garage. My first kiss. Your first kiss with me."
But that’s almost immediately followed by:
"I was the best cheerleading man in high school, and you were the best kisser in college. But you never kissed me glumly, always raising your hand majestically, like a majestically colored bikini. You kissed me so high that my boots creased the carpet beneath you."
"This town doesn’t cook delicious, but it does let you eat whatever the fuck you want. Your ovaries are like television, only they are hooked on the news. You can’t even spell rhyme or reason correctly. The stagehands send out the correct notes. The comedian sets the tone; the crowd gets funnier. It’s all a matter of perspective."
For a tone poem, one or two high points might be enough, but this thing is also at least two or three times longer than it should be – once my brain twigged to the fact that it was just watching a slot machine, I found it really hard to push on.
I do get the appeal, if not fascination, this sort of thing can have for a writer: the prospect of looking at your work through a glass, darkly, so you can apprehend it in a new way. I wrote a (very bad) novel in my younger days, about high-school wrestlers who I must confess were far more foul-mouthed even than Tom Trundle, and when I finished the first draft I excitedly ran it through Word’s (I think now deprecated) autosummarize feature, only to see my 300-page novel turned into three paragraphs of this:
"Fuck. “What if Coach found out?” Jesus, Coach. The first match I wrestled? Fuck. Three matches. Fuck. Fuck. “Yeah.” “Right.” “Yeah, right.” “Oh, right.” “Right.” “Redford?” “Yeah, yeah. Fucking math. Shit.” “Fuck. Fuck. Fucking test. Very helpful, Redford.” Redford nods. “Right.” “Fuck.” “Ah,” sighs Redford. “If only.” Everybody likes Redford. Anais? Fuck. “Yeah.” “Uh, right.” Redford’s word."
I died laughing – this is actually my ghost writing this – and resolved to tone down the profanity somewhat. As a way to change perspective, or add a bit of surprising flavor to the hand-crafted sauce (this metaphor has gotten away from me), I can see the value of procedurally-generated text – and I’m sure in twenty or thirty years Skynet will have come for the writers the same way it’ll have come for everybody else – but in the meantime, the werewolf’s getting no thank-yous from me.
So I had some thoughts that I was going to write here when I was about 60% of the way through the piece, but when I finished I read the About section and now I'm not sure any of it applies. The author notes that the majority of the text for this game was written by an algorithm that was seeded with some of the author's writing from the past 20 years. The result isn't so much surrealist as it is non-sensical and the choices you are offered don't steer the story, but rather just pick the next random piece of text for you to read. There isn't any story that I could discover in it. It is just a Jackson Pollock painting of words.
Clean execution. It is also accompanied by some nice music to further deepen the acid trip feeling.
**You Will Thank Me as Fast as You Thank a Werewolf by BJ Best**
I beta tested this game.
In regards to the scenery and trimmings of this game, it's polished and nice-looking in the Chapbook format or lookalike (I remember I asked about the format while testing but can't find my correspondence anywhere). It has good music and flows naturally.
Writing-wise, this is GPT-2 (a procedural generation/ai tool). I usually really dislike GTP-2 because it just regurgitates whatever's put into it. Most popular uses of GPT-2 involves scraping other people's content without attribution and then spitting it out, with most of the 'best results' being word-for-word copies of the original input.
But in this case, the person using GPT-2 is the person who made the original content, so that makes it more interesting. I guess, then, that this is like a procedurally generated mirror. It lets the author see themself, and it lets us see that vicariously.
There are fun parts in the writing (the line 'You count the days until Christmas. I count the days when we didn’t know each other’s last names.' reminds me of Arcade Fire lyrics). Overall, it's an interesting experiment, and reveals a lot about the author.
+Polish: The game is smooth.
-Descriptiveness: It's made of interesting chunks, but they don't flow together in larger picture.
+Interactivity: It gives the sense of interaction, a weird sense of pseudo-agency. The footnotes help.
+Emotional impact: For me it was curiosity about the author themself.
-Would I play again? No, one run through seems enough for me.
I enjoyed this, and was content to let the music wash over me, and the associations cohere and dissolve as I clicked through the story. It reminded me of my weirdo MFA program, which I really enjoyed. There is a flatness or generic-ness to the references--fields, tornadoes, school, parents, tv, birds, God, cigarettes--that became interesting when they were juxtaposed in such odd ways. I did find myself wanting a little more variety or tension. I don't know anything about GPT-2, but I wonder what would happen if different seeds or corpora (like with different styles of writing, or period-specific cultural references, or references to a specific place) were used within the same piece to modulate the tone a little bit more.
From its in-game About section:
“This work is a collaboration with GPT-2, a neural network model designed to predict the next word in a block of given text based on its study of eight million web pages. In this application, I input a text file of my own prose from the past twenty years into GPT-2. It then generated new writing in a similar style. I selected, arranged, and lightly edited the resulting output.”
I’d be entertained if somebody collaborated with GPT-2 to generate a review for this entry, but on the IFcomp rating scale of 1 to 10, I’m giving it a 2.