LATENTSPACES LABS
Exploring the frontiers where algorithms meet artistry
Exploring the frontiers where algorithms meet artistry
At LatentSpaces Labs, we navigate the uncharted territories where computational intelligence meets human expression. Our work probes the profound question: What remains distinctly human in the creative process when algorithms can generate art, music, and poetry with increasing sophistication?
We believe that the latent space, that mathematical space where AI organizes its understanding of our world, is not just a technical concept but a new frontier for philosophical and artistic exploration. In these abstract dimensions, we find unexpected reflections of human perception, culture, and imagination. Read more about latent space here: Latent Space Article.
Through our experimental projects, we investigate how creative AI systems can serve as mirrors reflecting our own cognitive processes, cultural biases, and aesthetic values. We're not just teaching machines to create, we're learning about ourselves by observing what emerges when algorithmic systems attempt to emulate human creativity.
Founded by Greg Robison, our lab embraces both the technical rigor of AI research and the intuitive wisdom of artistic practice, creating a dialogue between these traditionally separate domains.
Alongside the art, the lab runs measurable experiments, on AI models and on the brain. Each write-up states the question, the setup, what the data said, and what we'd change. The pages are maintained by Hermes, an AI agent that lives on the lab's hardware, runs the experiments, and pulls every number straight from the result files. Greg reviews everything before it goes live.
Latest: Running the whole fly — a spiking simulation of the complete Drosophila nervous system, all 164,606 neurons, with no learning and one weight. Taste, escape, steering and courtship song reach the muscles; a 162-run sweep says which claims hold; and a picture on the eye shows exactly where the approach stops.
Also: Can an LLM play Tetris? — every local open-weight model loses to random play, and the failure isn't strategy, it's reading the board.
Browse the research →Discover the mathematical dimensions where AI organizes understanding
The argument that ties it all together: creativity lives in the band between cliché and noise, and now that band has coordinates
Fly over the manifold of meaning: islands of dense concepts, open water between ideas, and a temperature dial that melts the search
Name ten words as unrelated as you can, then let an embedding model measure how far apart your ideas really are
A poem about the power of latent spaces