> The information can be anything. A seemingly uncoordinated twitching finger could be a song when placed on a piano or the right pin number when placed over an input panel of a cash dispenser. Until the signal is not decoded, even the carrier is clueless about its meaning.<p>Is that true? I've never heard of 'muscle memory' being an actual memory like that...<p>---<p>However, I'd like to point out that there is in fact a well-established kind of memory which lets you use a human to carry a large amount of memory while being 'clueless' about not just its meaning, but what the encoding even is: visual <i>recognition memory</i>.<p>You can recognize whether you've seen an image before; given appropriate datasets, this allows encoding messages into a human's visual recognition memory without them being able to elicit this knowledge in the absence of the target images. This can be used for tricks like storing passwords, or multiplying 10-digit numbers: <a href="https://people.csail.mit.edu/andyd/rec_method.pdf" rel="nofollow">https://people.csail.mit.edu/andyd/rec_method.pdf</a> I calculate that you could feasibly store 5.8 kilobits per human: <a href="https://www.gwern.net/Spaced-repetition#fn61" rel="nofollow">https://www.gwern.net/Spaced-repetition#fn61</a>