Re: Uploading

Anders Sandberg (asa@nada.kth.se)
25 Aug 1998 13:56:25 +0200

Emmanuel Charpentier <emmanuel_charpentier@yahoo.com> writes:

> Duplicating the processing while keeping only one neural net should
> resolve that, only neurones triggered in the first process can link
> together, they don't link with the second process (the eventual
> chemicals are also duplicated).

Yes, but when you merge the two nets the temporal relationship between the experiences could become garbled. I'm not sure what the result would be, but I'm fairly certain that answering questions such as "What did you do last tuesday?" and "After fixing the ridgeways clenchiration, what did you do?" would be very hard to answer.

> And if experience tells us that a part
> of the brain is necessary for short term memory, I am not sure it
> really acts as a hard disk or RAM drive, cound't it be some kind of
> cerebellum? And if it acts as a memory drive, why not (again)
> duplicate it?

Because the really short term memory areas (especially the hippocampus) most likely experience even topology change during the day. It becomes very non-trivial to duplicate and merge them.

> Having different bodies but one memory could be a great and funny
> thing. I could learn different things at the same time, be at
> different places, get into different activities, hold group meetings
> with only me into it (would that be very enriching?), but still be
> only one person. Is that a step to omnipotence? :D

I don't know, but I would certainly want to try the experience (as I have described in my own take on it,
http://www.aleph.se/Trans/Global/Posthumanity/borg.html)

> Another upgrade to ourselves would be adding a sense of self
> inspection: looking inside our own brain and eventually changing it!

Yes, this would be very useful. A source level debugger, or something like the toolkits mentioned in Egan's _Diaspora_ (they even enabled simulation of how the mind would look after accepting certain modifications, making it possible to decide whether they were dangerous or not).

A first step would be something like

@InProceedings{Laakso98,

  author = 	 {Laakso, Aarre AND Cottrell, Garrison W.},
  title = 	 {How can I know what you think? Assessing representational similarity in neural systems},
  booktitle = 	 {Proceedings of the Twentieth Annual Cognitive Science Conference},
  year =	 1998,
  annote =	 {http://www-cse.ucsd.edu/users/gary/pubs/cogsci98.laakso.ps}
}

where neural networks can be compared in terms of their internal representations. This approach might be extended to a way of figuring out the structures of these representations, and then visualizing them.

> The funny thing is that this sense would be part of the net, and
> should then be able to look upon itself and redesign itself, I wonder
> if there should be any blind sight corners in its capacities?

Most likely; we don't have full access to our brain, and a lot of stuff likely resides as wague "sub-concepts", fragments and other neural junk that is hard to relate to anything but affect us.

-- 
-----------------------------------------------------------------------
Anders Sandberg                                      Towards Ascension!
asa@nada.kth.se                            http://www.nada.kth.se/~asa/
GCS/M/S/O d++ -p+ c++++ !l u+ e++ m++ s+/+ n--- h+/* f+ g+ w++ t+ r+ !y