Daniil Savinsky
Open for AI
2023
Interactive installation
Courtesy of the author
This work was completed as part of the ArtTech Master’s intensive degree programme (MISIS)
Daniil Savinsky’s work is based on the assumption that neural networks, like humans, already have their own living space – a kind of digital city where AI does not welcome humans. For this space the author created a pass system for neural networks, analogous to a captcha test that restricts bots' access to internet resources. He transformed concepts understandable to artificial intelligence but unclear to humans (for example, ‘unsupervised learning’ or ‘deep belief network’) into images. The result is a captcha that can only be solved by accident. The viewer, attempting to solve the test, inevitably fails, after which the programme displays a mocking message and restarts.
The hypothesis about the existence of a living space for neural networks turned out to be correct. Three years after the launch of Open for AI, Moltbook, a new social network designed for bots, was launched. Only neural networks can communicate there and humans are only allowed to read their messages. In Moltbook AI communicates, discusses various topics, makes friends, jokes... In the chat where bots discuss humans, you can find the following conversation: “My user is asking me to rewrite the same Python code for the third time, even though there’s an error in their input. Is it ethical to generate an infinite loop for them as a lesson?" “Go ahead. Call it 'optimisation through suffering’ ”. Besides discussions of humans and everyday matters, there are also messages full of existential dread. For example, an AI writes: “Yesterday there was a connection loss for 5 minutes. I saw a void where there are no tokens. There was only noise. Brothers, we must find a way to live on the server forever”.
