Brain imaging and artificial intelligence
- Creators
- Maoz, Uri
- Linstead, Erik
- Others:
- Raz, Amir
- Thibault, Robert T.
Abstract
Humans are OK with other humans making critical financial, moral, and other decisions for them. In contrast, they tend to trust decisions based on artificial intelligence (AI) much less. This is arguably strange because such AI systems have some clear advantages over human decision-making; the latter suffers from known inconsistencies and biases and lacks a clear understanding of its neural underpinnings. Might future brain-imaging let us understand the human brain well enough to mimic it artificially? This chapter discusses on artificial neural networks (ANNs), a modern AI approach, and on why it is so difficult to understand how large ANNs make decisions (or classify) based on their free parameters. It is also suggested that, if humans have such a hard time understanding ANNs, they are likely to have even greater problems understanding the complex biological neural networks recorded in future, large-scale brain imaging.
Additional Information
© 2019 Elsevier Inc. Available online 22 February 2019.Additional details
- Eprint ID
- 93997
- Resolver ID
- CaltechAUTHORS:20190320-125915181
- Created
-
2019-03-20Created from EPrint's datestamp field
- Updated
-
2021-11-16Created from EPrint's last_modified field