The structure of evolved representations across different substrates for artificial intelligence
This addresses the problem of AI robustness to noise for researchers and practitioners, though it appears incremental as it compares existing methods without major breakthroughs.
The paper compared how different AI substrates (recurrent ANNs, LSTMs, and Markov Brains) represent environmental information, finding that Markov Brains localize and sparsely distribute information, while neural networks spread it across nodes, making them more vulnerable to noise.
Artificial neural networks (ANNs), while exceptionally useful for classification, are vulnerable to misdirection. Small amounts of noise can significantly affect their ability to correctly complete a task. Instead of generalizing concepts, ANNs seem to focus on surface statistical regularities in a given task. Here we compare how recurrent artificial neural networks, long short-term memory units, and Markov Brains sense and remember their environments. We show that information in Markov Brains is localized and sparsely distributed, while the other neural network substrates "smear" information about the environment across all nodes, which makes them vulnerable to noise.