Our various human senses do a very good job of complementing each other. Our sense of sight provides enormous bandwidth, while our hearing allows us to detect and locate events all around us. Touch lets us accurately assess the texture and solidity of objects, and permits us to use our amazing fingers and hands to manipulate tools.

One difference between our respective senses is in the scale of time at which each operates. For example, to simulate our sense of vision accurately, you need to flash about 100 different images in front of the eyes every second. Of course, that’s just a rough approximation, but it’s reasonably accurate. For example, 30 images per second isn’t quite enough, whereas 300 images per second would be overkill.

In contrast, to simulate our sense of touch you need to provide about 1000 changes in haptic sensation per second. If you only provide 300, then the things you are simulating will always feel spongy and soft, whereas 3000 haptic sensations per second feels pretty much the same as 1000.

To create a perceptually perfect simulation of sound, you need to go up to about 100,000 vibrations per second. If you try to get away with 30,000, you end up losing phase information at the high frequencies, which makes some objects sound like they aren’t quite coming from the right location. On the other hand, anything above 100,000 vibrations per second would pretty much be pointless.

I find it interesting that we’ve got these three nice round numbers: 100, 1000 and 100,000, for visual, haptic and auditory “samples per second”, respectively.

So what happened to 10000? We seem to have skipped right over that one. I’m pretty sure we won’t find it in our sense of smell, since smell works at a very slow time scale, compared with most of our other senses.

Maybe in the future, after we have figured out a practical way to interface directly to the brain, we will discover that some high-tech version of mind reading requires 10000 neural samples per second.

But it might be a while before we know whether that is true or not. By my reckoning, we will probably need to wait about seven trillion neural samples, more or less, before we find out.

Leave a Reply