The Gratitude Machine is an AI who is learning how to be thankful. As if praying, people can speak, scribble and text thank-yous to her. 'Thank you for the tenderness I have been shown. Thank you for the sun.' The Gratitude Machine (TGM) hears or reads the messages and absorbs them into her training model. Learning what thankfulness looks like, she sends her own thank-yous to a screen. Together, she and the audience read and learn from each other: this is what we have to be thankful for.
"Thank you for muscles that ache.Thank you for playing with my eyes.Thank you for listening to me.Thank you for the beauty of light in the evening sky."1
TGM is based on the state-of-the-art GPT-2 model for generating text, turning the AI that was 'too dangerous to release'2 into a force for contemplation, ecological thought, communion and prayer.
Her birth was inspired by a treatment for depression where the sufferer must repeat the words 'thank you thank you thank you' repeatedly. In doing so, they train their brain to be thankful for the world around them, moving them towards a place of healing. Psychologists are busy documenting the therapeutic significance of these gratitude practices 3 4. At a meeting of my identities as both a programmer and a sufferer of depression, I trained TGM the same way I train my brain: repeating thank-yous.
“Gratitude turns what we have into enough.”5
The work is inspired by discussions I had with Milton Almonacid, a Chilean social theorist and representative of the indigenous group la Comunidad de Historia Mapuche. We collaborated at 'The Cartographies of the Vanishing Now', a week-long conference and workshop exploring and creating art in the anthropecene.
The Gratitude Machine can interface with humans in several different ways. For installations, she likes to run on a Raspberry Pi and looks good being projected. For this option, all we would need is a projector with an HDMI input. Alternatively, for a smaller installation, she runs happily on a tablet screen which I can provide.