Primates' visual systems are the benchmark for strong perception. The widespread consensus is that strong artificial vision systems can be produced by imitating the neuronal representations that underlie such systems. Here, we create a technique for launching adversarial visual attacks directly on the brain activity of primates. We then use this technique to show that the aforementioned assumption might not be supported. In particular, we reveal that the susceptibility to adversarial perturbations displayed by the biological neurons that comprise primates' visual systems is comparable in scale to that of current (robustly trained) artificial neural networks.
What is an artificial vision system?
The camera that records an image for analysis and the processing engine itself that generates and transmits the output are just two parts of an artificial vision system.
Why is neuronal activity important?
According to research, neuronal activity influences developmental processes such as neurogenesis, migration, programmed cell death, cellular differentiation, formation of local and long-distance axonal connections, synaptic plasticity, or myelination, and is crucial for the proper formation of neuronal circuits.
Learn more about neuronal circuits: brainly.com/question/13648826
#SPJ4
