Re: Uploading: End of the mundane

Hal Finney (hal@rain.org)
Wed, 19 Aug 1998 21:56:52 -0700

James Millar, <jamillar@socs.uts.edu.au>, writes:
> Why not download the "lessons learned" experience directly into your mind
> from the copies mind. This would remove the need for reading a summary
> and/or repeating tasks.

Scott Badger, <wbadger@psyberlink.net>, writes:
> My thoughts exactly. Program the copies to self-destruct in 8 (or whatever)
> hours and assimilate the information into your consciousness. In fact,
> program thousands of short-lived copies and assimilate their experiences and
> you got your own personal singularity going.
>
> I also have a hard time conceptualizing the notion of an uploaded mind
> desiring a vacation. Would we *need* vacations in that state? I mean
> vacations are basically a two-week reward for spending 50 weeks at something
> you don't particularly like to do. As uploads, aren't we likely to
> experience a greater level of work satisfaction? I mean, how important
> would vacations be if you *really* loved your job?

You both seem to be assuming that uploading gives more abilities than simply the ability to simulate a brain on the computer: the ability to download experiences directly into the simulated brain, the ability to automatically assimilate information, the ability to adjust brain state so that work is rewarding.

While it is conceivable that by the time we are able to upload, we will understand the brain well enough to have these abilities, it is not necessarily true. Uploading may be as simple as being able to simulate how individual neurons work, and having the ability to scan the brain at the neural level. This could be enough to scan in a brain and simulate it accurately so that it behaves just like the real brain. If this route is followed it will not automatically inform us about how to insert new experiences into the brain, or what to tweak in order to change rewards.

Actually, the technology needed to do this kind of scanning might well let us manipulate physical brains almost as easily as we do virtual ones. Perhaps we could crudely solve the reward problem by stimulating the "pleasure center" whenever the desired activity occurs. This could probably be done in both simulated and real brains.

Assimilating experiences will require learning a great deal about how memory works. Experiences go into memory and somehow get associated with other, similar memories. It might be that the only way to know how a given brain would react to a certain experience would be to actually simulate that brain's exposure to the experience.

Suppose I am an upload and I buy a program which will insert the knowledge of French into my brain. The program makes a copy of my brain, suspending the original. It then laboriously teaches the copy French, by repetitive drill, practice, etc., just as it is done today. This is how it finds out exactly how my particular brain would represent the knowledge of French.

At the end of this process it copies the brain state from the copy back into the original, destroys the copy, and resumes the original. Presto, knowledge of French has been painlessly "inserted" into my brain. But it is obvious, I hope, that this process is effectivelly the same as having me learn French the old fashioned way.

Hal