Kurzweil uploads

[jordansparks]

Kurzweil's latest book just came out. I'm not going to discuss his AI predictions because we each have our own constantly evolving opinions about the trajectory of AI. But what caught my attention was the idea of using nanobots in the brain capillaries to perform a mind upload. I've always said that nanobots would be useless in reviving cryonics patients, and I still feel that way. But nanobots in the blood would be quite a different beast. These nanobots would not have arms that manipulate or repair molecules inside cells. They would instead be restricted to the capillaries. Any probe appendages would not be used for repair but would instead be used to sense and monitor the electrical flow of brain activity. That's very different and it's much easier -- still not happening in 20 years, but much much easier.

My previous upload scenarios generally involved slicing and dicing the brain into many millions of pieces which then get scanned with electron microscopes. But I now think nanobots in the brain capillaries is an equally likely scenario, so I will be working that into my https://oregoncryo.com/futureTechnology.html page. I completely disagree with Kurzweil's timeframe of 20 years, of course. 50 years for the first extremely crude capillary nanobots that would be widely available to the public is more more realistic, with 90 years until they would be capable of a full mind upload. In other words, I predict it will be possible at about the same time as the scanning approach. But there are some huge advantages to the nanobot approach. First of all, we would get that 40 year window where we would be connected directly to the cloud without being able to upload. This would make it a gradually improving iterative process. This would obviously help the public understand that uploading would eventually be possible. It would also have many fantastic short term benefits separate from uploading, which means that vast resources would be poured into that technology. Because of the gradual nature of the improvements, nanobots feels a bit more likely than scanning.

I'm on board with the nanobot scenario now, but not the kind that repair anything. I think most nanobots will be passive monitors or will perform crude and simple tasks. They will obviously start in the gut instead of the capillaries. 25 years to the first widely available gut robots is not unrealistic. Unfortunately, my estimate of 50 years to the first capillary robots means I will completely miss the boat. But maybe it could still be fun if someone comes up with a robot that lives in my gut and runs on sugar or something cool like that. My guess is that the form factor for that would look like a 6" worm. Thank goodness I have a weak gag reflex and would be able to swallow one of those every few weeks.

[PCmorphy72]

I would add a 30 June 2044 event on my Google calendar in order to properly continue this discussion and verify if Kurzweil was too optimistic or not. If you will still be so determined perhaps your forum will survive to the Google accounts as we know today, but the will of the people involved in tecnological progress is erratic when there is not a wide interest in the community, that's why surgeons and sonographers still prefer their hands to the robots (and why gastroenterologists exist to prefer "direct ways" to monitor our guts, in order to win in an easy way against our abdominal muscolar resistance: after all they would deal with a brain as it was a gut).
I remember when I saw the Star Trek 1979 movie in the eighties I was sure that in 2000 medical monitoring would have been something similar, but... Making shapshots (scans) is a more confortable way of development than supplying intelligence (artificial or not) to robots (nano or not), when the intelligence is not used only for stupid demonstrations: that's true since the eighties. More intelligence has certainly been used for nuclear fusion or quantum electronics. Miniaturization, until now, has served the community mainly for calculation purposes, not for medical monitoring. By the way, did Kurzweil predict also visual monitoring by the nanobots, apart "the electrical flow of brain activity"?

[jordansparks]

The beginnings of this technology already exist. Stentrode is a stent that gets implanted in blood vessels in the brain. It can be used in patients with severe paralysis caused by stroke, injury, ALS, etc. The stent has electrodes on it, making it a brain-computer interface. This feels like a much better approach than cortical surface electrodes like Neuralink. No cutting a hole in the skull, no array of wires piercing the brain and causing damage, indefinite functionality instead of planned failure, and so on. Stentrode is currently implanted in four humans and helps translate their thoughts into digital output. The implants have been successful, with patients being able to use it to control a mouse unsupervised at home.

It seems obvious that this approach is good. I suspect that Stentrode use will become widespread over the next few decades. It will take quite a bit of imagination to predict how those electrode stents could get smaller and more numerous. If I imagine us developing very good microrobotics, then the way that would be put to use would be to make the placement process more refined, kind of like how vertical oil well drilling morphed into being able to steer the drill and go sideways. So the surgeon implanting these brain stents would end up with really good robotic tools. This would enable placing stents into many smaller veins instead of just the superior sagittal sinus. The next improvement I can imagine would be wireless, but the moment you go wireless, you need to worry most about getting power. Maybe the power source could be in the larger vessels, and the devices in the smaller vessels get their power from induction. Or maybe patients would need to wear an induction helmet to periodically recharge the onboard batteries of these stents. Long term, it seems like they will probably need to run on sugar. In any case, once they solve the wireless problem, the number of stents increases. It feels like that's about the time when the general healthy public will start to want them. If we have that level of technology, then we will already have had decades of experience with robots existing inside us. And that's also about the same time that we will finally have cloned replacement organs available, so routine heart replacement will be widespread. That's about 60 years from now, and it would still take quite a bit of additional progress before we could end up with tens of thousands of these brain stents in each of us to allow us to get closer to upload capability.

[JohnMSmart]

This is a fascinating idea Jordan. If capillary nanobots could both read electric fields and generate fields, and they each had unique IDs and could communicate wirelessly with external machines, and if their position could stay fixed for life, they might become a very high resolution brain-computer interface (BCI). This could allow stable telepathic links between biological brains and their technological brains and bodies, and lots of new capabilities for people.

I'm doubtful that such a BCI could be a great uploading device however. I can see that the capillary-based sensor mesh could capture large amounts of data on dynamic activity in response to recorded internal and external activities. But it wouldn't be a model specific to individual neurons, and it wouldn't have access to the unique details of each neuron's synaptome. One might infer circuit structure to some degree, but I'd expect this wouldn't capture many of the fine details of our individual identity. Perhaps I'm missing something in this analysis though. Thanks for the post and the Stentrode update too. Great to see that progress occurring!

[jordansparks]

Any repair or upload technology will involve massive inference. Also, with implants, I'm not suggesting uploading until the technology is very mature. Since we will have had decades of experience with these implants by then, I suspect there will come a point when the pros of uploading will outweigh the cons of losing some information. For example, the meat brain will continue to deteriorate with old age. Maybe those people will have no choice but to upload if they want to maintain continuity instead of sitting out a few years in stasis. Maybe by that time, the amount of external computing power will be vastly greater than in the meat brain. And maybe the speed will be much higher. Remember that we only compute at about 10 bits per second. So there could be many reasons why people are willing to make that jump.