About Emiliano's Lab
Emiliano's Lab is a research lab built by Emiliano Cuevas. The aim is to understand brain and intelligence at different levels and scales: predicted cortical responses, measured EEG, language and perception, and evolutionary simulations of adaptive behavior. Every study poses a clear question, uses appropriate data or models, and interprets results carefully.
Computational & in-silico studies
Several experiments use TRIBE v2, a multimodal brain encoding model from Meta Research. It was trained on over 1,000 hours of neuroimaging data across 720 subjects and can predict the fMRI response of the human cortex to text, audio, and video inputs. Those projects use the “unseen subject” mode, which returns group-average predictions rather than individual brain responses — useful for asking what a model says about condition differences without collecting new fMRI.
Empirical recordings
Other experiments use data measured directly from participants — such as EEG — with standard acquisition, preprocessing, and analysis pipelines. Those write-ups focus on design, signal processing, and inference for that modality.
Evolution & origins of intelligence
Some work steps back from cortex and scalp signals to ask how adaptive behavior can arise under selection alone. Origins is a minimal evolutionary simulation where organisms carry a full joint distribution over local sensing, internal energy, and action — offspring inherit mutated joints, and survival in a crowded ring world provides the only fitness pressure. These studies complement encoding and EEG work by probing intelligence from the bottom up.
What does “in silico” mean?
In silico is Latin for “in silicon” — experiments performed on a computer, as opposed to in vitro (in glass) or in vivo (in living organisms). Here, it usually means predicting or simulating brain activity computationally rather than measuring it in a scanner, though the lab also includes in vivo recording work.
How the experiments are organized
Encoding-model studies often follow a fixed pattern: feed stimuli into an encoding model to obtain predicted cortical maps (for TRIBE v2, 20,484 vertices per sample), then train a classifier or run statistics to test whether representations differ across conditions. EEG studies follow acquisition and preprocessing suited to the hypothesis. Simulation work evolves populations under explicit world dynamics and logs distributional metrics over time. Each experiment page documents its own methods and limitations.
Built by Emiliano Cuevas — a research prototype, not for clinical use.