[Continuing from the previous entry.]
As I drifted to sleep that night, I couldn't shake the feeling that I was still connected: my brain was reproducing the feeling, just like when you close your eyes after a long day of hiking and still see the trail in front of you. I was seeing different chess positions, and felt her reactions buzzing in the background. Even the alien abstract thoughts were being reproduced, rolling around and snapping chaotically. I felt like my brain was trying to make sense of them as I lay there in the dark.
The next day, I took out the chess set again and played a game against myself. I knew I wasn't playing nearly as well as when I was connected to Chrissy, but I thought that I wasn't playing like myself, either. I hadn't totally lost the chess instincts that I stole from her.
If we played chess a few more times, I could see that I would be at her level even helmet-free... at least if I kept practicing. And she wasn't even trying to teach me-- in fact, she had tried hard not to give me all her skills. With a good teacher, the helmet would be far more effective at transferring complex skills than any traditional method.
I skipped my classes that day, spending the time writing a primitive skill-transfer program for the helmet. The basic idea was clear: record expert brain activity at a task, and then stimulate the same activity in a novice. Unfortunately, to do it well, you would need sophisticated matching techniques to determine when it's appropriate to nudge activity in a particular direction. The best way of doing it would involve detailed mind-to-mind translation dictionaries identifying what activity patterns represent "the same" concept. I was still using the simple literal translation technique I had used with Chrissy any my other sync experiments.
I made a few recordings of myself drawing different things. I wasn't a really good artist, but I wasn't bad either, and I thought that maybe the motor control needed for drawing would be simple enough to transfer from one person to another. Ideally, the software would learn the connection between visual imagination and motor control, and then another person could run this, the software hinting at what lines to draw when they imagined what they wanted to make a picture of. My one-day effort had not produced anything this sophisticated. Instead, you could play through the memories of me drawing two different things (a face and a house, both of which took me about an hour) and hopefully learn something. I only included recordings of a few brain regions, though: I certainly didn't want to be posting all my thoughts online. The visuals, motor control, and sensations from the hand would have to be enough.
Other people were posting recordings of things like dreams, so my software wasn't really anything new at this point-- though I did include some basic features which would make it better than the normal EMNSA playback mode. The speed of the playback would attempt to adjust based on how well a person was syncing to it, and the recording would stop if you got too far off track, allowing you to restart a few seconds back to try again.
I posted it on a popular EMNSA forum and went to bed.
It was a few days before I saw Chrissy again, because of my class schedule.
How are you going to keep naming these?
ReplyDeleteOne idea I had that I forgot to comment on last time: the device is trying to sync without making detailed inferences about correct correspondence, so it is probably throwing away some information (or merely representing it in ways the receiving brain can't actually capture, eg mapping two neurons to one). One naive strategy for allowing more complete telepathy would simply be adding 'texture' to the signal; making sure that some consistent correspondence was available to the receiving mind for any potentially meaningful difference in the sending mind. This would be a little like a hash since you want it to be random but consistent given a particular sender and receiver; yet, since peoples' brains will shift over time you would potentially want the process to be fairly continuous... actually besides that, similar signals from the sender should have pretty similar 'texture' added for the receiver, when possible. So it might be a bit of a challenge to write and test.
Of course some detail must be lost, if only because the brains do not match up at the smallest level. I can see how forcing some precise mapping, accepting distortion rather than loss, could be beneficial. If that's the idea, the hash may not be difficult to generate. I'm imagining that a much bigger problem is simple image resolution, though. I guess the EMNSA I (we'll assume EMNSA II and so on would come out at a later date) has resolution somewhere between the centimeter and millimeter level; that's optimistic considering that it's supposed to be consumer tech, worse than research-level EEG. (State-of-the-art consumer EEG in real life has 14 sensors; research-level can have up to 256 sensors. Either way, not very close to millimeter level!) Transcranial magnetic stimulation has its own resolution problems, too, of course.
ReplyDelete