For curators
Practical information for assessing the installation. Where a figure is not yet certain we say so, and details are agreed with the venue.
Space
- Tape
- About 3 to 4 m long.
- Footprint
- A wall or clear surface somewhat longer than the tape, for the motor and the ends of the rail: roughly 4 to 5 m. Height and depth depend on the structure and will be set on the prototype.
- Light
- Steady light on the tape. The camera has to tell black from white reliably, so backlight and light that changes through the day should be avoided.
- Listening
- The clicks of the cells and the AI's voice are part of the work. A room that is not too noisy, or set apart from nearby sound works, is best.
- Minimal version
- A table and a power socket. See Technical.
Power
Standard power sockets near the installation, for the tape electronics, the computer and the speakers. Exact consumption will be measured on the prototype.
Duration
- Sessions
- A session runs until the program on the tape halts, or for a set number of steps or a set time. The machine is slow, so a session can be long. Sessions can follow a schedule.
- Continuous
- Sessions follow one another for the whole opening hours. Between sessions the tape returns to its initial state.
Network
A cloud language model needs a stable internet connection, preferably wired. If there is no network, or it is unreliable, there are two alternatives.
- Local LLM
- The model runs on the installation's computer with no connection. This needs a more powerful computer, and the answers may differ from the cloud model and come more slowly.
- Recorded mode
- The machine replays archived sessions: it executes the recorded steps and plays back the voice and story from that session. No model is running and no new errors happen.
Installation and support
Set-up time, transport and support during the exhibition are agreed with the venue.