Curiously, there seem to be people interested in actually doing this sort of thing.
For example, it is argued that one could never record enough sensory experiences and actions to produce brain emulation [Gemmell, Bell et al., 2006]. However, this is the right answer to the wrong question. The question is not whether one can replicate the 10 trillion synaptic strengths and yet greater number of connectivities of the human brain in a software matrix. This would be like trying to replicate human flight by building an airplane out of a trillion micro-widgets in the exact same configuration as found in an eagle or a sparrow. Instead, the goal here is to replicate the functionality of a specific human consciousness in software. There is no more reason to assume a priori that the only way to do this is to replicate a human brain than there is to assume a priori that the only way to fly is to replicate a bird. Instead, we reason that software emulation of a human mind via analysis of a set of mindfiles is achievable because there are but a limited number of personality traits [Costa & McCrae, 1990], a finite set of universal facial expressions and emotions [Brown, 1991], a diminished repertoire of remembered thoughts from day to day [Ebbinghaus, 1885], and not more than a few gigabytes of remembered information [Landauer, 1986].
In general, dozens (n) of mannerism, personality and feeling types (m) yield many thousands of unique human combinations via (n!)/(m!*(n-m)!). Once you add to these thousands of personality and worldview templates differential recollections, beliefs, attitudes and values (the gigabytes of remembered information) there are many billions of unique possible combinations of human psyches, one of which will be a best-match for digitally stored mindfiles on a predecessor biological person. Mindware best fits one of the “m” compound mannerisms, personality and feeling types to that analyzed from stored mindfiles, and then populates it with the recollections, beliefs, attitudes and values evident from the stored mindfiles.
From “The Terasem Mind Uploading Experiment”. Not sure how seriously to take them, but my own readings have been inclining me to the point of view that personal identities are just not that information-rich.
but my own readings have been inclining me to the point of view that personal identities are just not that information-rich.
Could you say more about said research? My sense is that that people can be flexible about preserving a tiny portion of their unique information, e.g. that many people would be very happy to forget most of their daily experiences from their lives so far (replacing with brief text files, records of major relationships and their emotional intensities, that sort of thing) in exchange for greatly extended life in the same body. But the “mindfile” backup method inevitably involves a chance for the original to diverge and evoke intuitions that “the thing over there, which I perceive as separate” doesn’t provide continuity.
It’s basically as above. Traits like IQ offer remarkable predictive power; Big Five on top of IQ allows more prediction, and the second paper’s Small 100 seems like it’d add nontrivial data if anyone runs it on a suitable database to establish what each trait does. Much of the remaining variation can be traced to the environment, which obviously doesn’t help in establishing that human personalities are extremely rich & complex.
Long-term memory is much smaller than most would guess when talking about ‘galaxies of galaxies of neurons’, and autobiographical memory is famously malleable and more symbolic than sensory. (Like dreams: they seem lifelike detailed and amazing computational feats, but if you try to actually test the detail, like read a book in your dream, you’ll usually fail.) Skills don’t involve very much personality, either, since there are so many ways to be a bad amateur and so few to be an expert (consider how few items it takes to make a decent expert system—not millions and billions!) and are measurable anyway. “The mind is flat”, one might say. Once you get past the (very difficult) tasks of perceiving and modeling the world and controlling our bodies, which we all have in common, is there really that much there?
More generally, people tend to think that they do things for complex and subtle reasons, while outsiders see them doing things for few reasons and transparent self-serving ones at that, IIRC; why do we believe the Inside view that we’re so very complex and unique special snowflakes, while ignoring our Outside view that everyone else seems to be fairly simple?
(More personally, I have more than once had the experience of reading something, composing a comment in my head, and going to make a comment—only to see that a past me had already posted that exact same comment. This is not conducive to thinking of myself as a complex unpredictable person, as opposed to a fairly simple predictable set of mechanisms.)
If you know of any essays or papers arguing something like the above but more rigorously, I’d appreciate pointers.
So for beta uploading, I would say as a general principle you want to go in order of variance explained: start with your whole genome (plus near relatives & tissue samples in order to catch rare/de novo variants & childhood influences), then use psychology tests & surveys designed for predictive power and precise measurement of important factors, then move on to recordings of interactive/adversarial material (like chat logs or speech), then record writings in general, then record quotidian details like your daily environment.
Curiously, there seem to be people interested in actually doing this sort of thing.
From “The Terasem Mind Uploading Experiment”. Not sure how seriously to take them, but my own readings have been inclining me to the point of view that personal identities are just not that information-rich.
Also relevant is another paper in that issue, on very large scale use of personality questionnaires: http://www.worldscinet.com/ijmc/04/0401/S1793843012400082.html
first link is broken, but available on the Wayback Machine: The Terasem Mind Uploading Experiment
second link is broken; was it linking to this article?: How Accurate Are Personality Tests?
Could you say more about said research? My sense is that that people can be flexible about preserving a tiny portion of their unique information, e.g. that many people would be very happy to forget most of their daily experiences from their lives so far (replacing with brief text files, records of major relationships and their emotional intensities, that sort of thing) in exchange for greatly extended life in the same body. But the “mindfile” backup method inevitably involves a chance for the original to diverge and evoke intuitions that “the thing over there, which I perceive as separate” doesn’t provide continuity.
It’s basically as above. Traits like IQ offer remarkable predictive power; Big Five on top of IQ allows more prediction, and the second paper’s Small 100 seems like it’d add nontrivial data if anyone runs it on a suitable database to establish what each trait does. Much of the remaining variation can be traced to the environment, which obviously doesn’t help in establishing that human personalities are extremely rich & complex.
Long-term memory is much smaller than most would guess when talking about ‘galaxies of galaxies of neurons’, and autobiographical memory is famously malleable and more symbolic than sensory. (Like dreams: they seem lifelike detailed and amazing computational feats, but if you try to actually test the detail, like read a book in your dream, you’ll usually fail.) Skills don’t involve very much personality, either, since there are so many ways to be a bad amateur and so few to be an expert (consider how few items it takes to make a decent expert system—not millions and billions!) and are measurable anyway. “The mind is flat”, one might say. Once you get past the (very difficult) tasks of perceiving and modeling the world and controlling our bodies, which we all have in common, is there really that much there?
More generally, people tend to think that they do things for complex and subtle reasons, while outsiders see them doing things for few reasons and transparent self-serving ones at that, IIRC; why do we believe the Inside view that we’re so very complex and unique special snowflakes, while ignoring our Outside view that everyone else seems to be fairly simple?
(More personally, I have more than once had the experience of reading something, composing a comment in my head, and going to make a comment—only to see that a past me had already posted that exact same comment. This is not conducive to thinking of myself as a complex unpredictable person, as opposed to a fairly simple predictable set of mechanisms.)
If you know of any essays or papers arguing something like the above but more rigorously, I’d appreciate pointers.
So for beta uploading, I would say as a general principle you want to go in order of variance explained: start with your whole genome (plus near relatives & tissue samples in order to catch rare/de novo variants & childhood influences), then use psychology tests & surveys designed for predictive power and precise measurement of important factors, then move on to recordings of interactive/adversarial material (like chat logs or speech), then record writings in general, then record quotidian details like your daily environment.