I argued in Nature Neuroscience in 2000 that the 'expertise hypothesis" seemed implausible because "the visual features that are diagnostic in discriminating cars (for example) are bound to be different from the features that are diagnostic in discriminating faces." Now @KathaDobs and @pranjulgupta have shown this rigorously with CNNs here:
https://mastodon.social/@KathaDobs/109711150946753646