52. Cognitive Futures (w/ Metahaven)
Show notes
We’re joined by Vinca Kruk and Daniel van der Velden of Metahaven, the Amsterdam-based artist collective known for their filmmaking, writing, and design. In some part we're here to talk about their new book, The Hard Question of Art: Cognitive Futures (Verso, 2026), though also at the top of our minds is Latent Spacecraft: Brains, GANs, Finnegans, written this year with Nina Beguš, Gašper Beguš and Riccardo Petrini for Antikythera.
The Hard Question of Art opens by conceding that art does not change the world, at least in the way we were told it would. For Daniel, art does its actual work by being perceived, and any politics of art has to start from what that perception does.
Latent Spacecraft starts from a GAN trained on eight words of English that begins inventing new ones the way a child does, and lets you walk around inside its vector space. On a walk Daniel found a battered Finnegans Wake on a free sidewalk bookshelf, 30 or 40 words on page one that appear in no dictionary, and a GAN trained on Joyce babbles Joycean, which Whisper and GPT-4 collapse into different subtitles. Daniel brings in Anthropic’s J-space paper, where researchers found words Claude has in mind before it answers, as the case for treating a model’s interior as a place you can enter.
We're so delighted to have them on the pod!
The Hard Question of Art opens by conceding that art does not change the world, at least in the way we were told it would. For Daniel, art does its actual work by being perceived, and any politics of art has to start from what that perception does.
Latent Spacecraft starts from a GAN trained on eight words of English that begins inventing new ones the way a child does, and lets you walk around inside its vector space. On a walk Daniel found a battered Finnegans Wake on a free sidewalk bookshelf, 30 or 40 words on page one that appear in no dictionary, and a GAN trained on Joyce babbles Joycean, which Whisper and GPT-4 collapse into different subtitles. Daniel brings in Anthropic’s J-space paper, where researchers found words Claude has in mind before it answers, as the case for treating a model’s interior as a place you can enter.
We're so delighted to have them on the pod!