Acting rationally

Generating intelligent behavior

Thinking humanly

Acting humanly

The machine’s goal is to predict a concept or word tied to specific brain activation patterns.

The analyzed fMRI images included the bone structure of each subject’s skull.

fMRI images could be replaced by cheaper scans (e.g., X-rays or CT scans) for this experiment.

fMRI images can be used to train a computer to identify subjects’ intentions.

Deep Learning (or Classical techniques)

Perception and Localization

Motion Controllers

Behavior Arbitration

Decreasing model size and complexity

Advancing toward Artificial General Intelligence

Reducing model multi-modality

Remaining within the current scope of abilities

Translating languages automatically

Identifying the mood or emotional tone behind a series of words

Detecting network intrusions

Correcting grammatical errors in text

By using faster engines

By using alternative fuel

By only operating on highways

By communicating and coordinating with other vehicles

The complexity of the brain's neural network.

The physical structure of the brain.

The voxel pattern of brain activity.

The label associated with the image inducing the brain activity.

Word

Phoneme

Utterance

Morpheme

Neural networks

Reinforcement learning

Unsupervised learning

Supervised learning