Engineering a Less Artificial Intelligence.
By Fabian H Sinz, Xaq Pitkow, Jacob Reimer, Matthias Bethge, Andreas S. Tolias. 2019
168 citations recorded by the source corpus.
Despite enormous progress in machine learning, artificial neural networks still lag behind brains in their ability to generalize to new situations. Given identical training data, differences in generalization are caused by many defining features of a learning algorithm, such as network architecture and learning rule. Their joint effect, called "inductive bias," determines how well any learning algorithm-or brain-generalizes: robust generalization needs good inductive biases. Artificial networks use rather nonspecific biases and often latch onto patterns that are only informative about the statistics of the training data but may not generalize to different scenarios. Brains, on the other hand, generalize across comparatively drastic changes in the sensory input all the time. We highlight some shortcomings of state-of-the-art learning algorithms compared to biological brains and discuss several ideas about how neuroscience can guide the quest for better inductive biases by providing useful constraints on representations and network architecture.
Subfields: Artificial Intelligence
How researchers see this paper
The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.
In Artificial Intelligence
- Matteo Negri
- Marco Turchi
- Christopher Summerfield
- Stephanie Nelli
- Marco Gaido
- Beatrice Savoldi
- Luisa Bentivogli
- Andrew M. Saxe
Citations in the corpus
- If deep learning is the answer, what is the question? (2020, background): “The recent literature is studded with comparisons of the behaviour and activity of biological and artificial systems 7–21 , summarized in a growing number of review articles 22–30 .”
- A massive 7T fMRI dataset to bridge cognitive neuroscience and artificial intelligence (2021, background): “…to task-optimized networks 16,20 , as the narrowly defined tasks that task-optimized networks are typically trained to solve do not necessarily respect the diversity of functions supported by the human visual system 49 nor necessarily match properties found in biological visual systems 50 .”
- Gender Bias in Machine Translation (2021, background): “What is ‘‘easy to learn’’ for a model depends on the inductive bias (Sinz et al., 2019; Geirhos et al., 2020) resulting from architectural choices, training data and learning rules.”