The most damned achievement of science.
By the beginning of 2023, everyone seemed to have heard about neural networks: algorithms had learned to imitate human speech, generate drawings and photographs, and even successfully write graduation theses. Neural networks are capable of performing such a variety of tasks that they can already be entrusted with complex projects.
User watchmeforever came up with perhaps the most ambitious task for neural networks: filming the endless sitcom Nothing, Forever (“Nothing, always”). This, so to speak, art object has been continuously broadcast for more than a day — and this is the strangest, creepiest and wrong thing I have seen on the platform.
How does it even work?
The author used several neural networks for different components of the series. Apparently, he initially generated the sets and character models, and also took several ready-made tracks with voice-overs — this is the only thing that remains unchanged.
The scenario of an endless sitcom is being invented on the go by the OpenAI’s GPT-3 neural network — the very sensational chat bot capable of writing articles, dissertations and even pieces of working code in different languages.
It is not known exactly what requests are sent to the bot, but from what is happening, it endlessly generates scenarios for scenes from an old-fashioned sitcom like Seinfeld.
The resulting text is acted out by a created set of characters: they voice lines with synthetic voices and move around the room. Sometimes, off-screen laughter sounds in the background — usually completely out of place.
Is it watchable?
If you evaluate Nothing, Forever as a regular series — this is trash in the worst sense. In the series there is no plot development or even a conditional logical connection between the scenes, there is no drama, there are practically no at least relatively funny jokes. And still, few real shows can evoke such strong emotions. To be more precise, a feeling of chthonic horror.
Nothing, Forever is an absolutely amazing art project:
- This is the peak of absurdity — not a single living person could write something so illogical and meaningless.
- The series is funny just because it does not make sense: sometimes what is happening breaks into laughter because of the cringe.
- Blatant errors of neural networks are also funny: for example, sometimes characters voice script lines that should be instructions. For example: «Pause for laughter», «He walks strangely around the room.» Sometimes the camera simply does not reach the heroes, and they themselves cannot even sit on a chair without breaking their necks or falling through the textures.
- It is full of the most awkward pauses — the characters can sit in silence even the entire scene.
- Neural networks cannot even roughly imitate dramaturgy: this is one endless plotless episode, the scenes in which end in the middle of the dialogue.
- The series becomes like a real show only in those rare cases when the neural network produces practically unedited excerpts from real scenarios and anecdotes.
- If you watch the stream with the chat turned on, Nothing, Forever acquires a human context and becomes more interesting: the audience reacts violently to every “joke”, invents memes and even slowly takes a liking to the characters.
Great Dialogues from the Series
Today I was walking down the street and I met a pigeon. It took off so unexpectedly that I almost dropped my coffee.
— Wow, that sounds GREAT!
Yes, it was funny to see the dove so quickly.
“I heard that they find a mate for life. Maybe he was looking for a partner?
“That would explain a lot.
– My God, have you seen this new place? It’s like straight out of a sitcom.
*offscreen laugh*
— Yes I know. Could be the setting for a burger joint.
Ha, I doubt it! Larry would never let us in.
*laughter*
I think we’ll spend the night in jail. And the deputy would not praise us.
“Don’t you think this place is too posh for us?”
“Yeah, I think we should spend time in a pretty dirty spoon.”
* * *
Seriously, Nothing, Forever is the best proof that neural networks are still in their infancy and do not yet understand things like context and logic, but only imitate them based on given patterns. So, to create real artificial intelligence, developers have to do many times more work.
