fafromnice wrote:

If i want to build an Margareth Tatcher AGI and I hate Margareth Tatcher I will program the AGI with racist, sexist, war monger Tought Paterns. In the other way around if I like Tatcher it's possible that I will program her/it with a mother-ly view for his/her people

Yes; but let’s assume there’s some kind of competitive ethic about building simulative personalities that’s centered around accurately representing that persona—otherwise, what’s the point? A phony product would be perceived as shoddy and widely ridiculed. Thinking about this, there are perhaps two (probably more) independent and therefore complementary means to approach accurately representing a generated personality: 1. Assuming a digitized mind carries some sort of pattern—let’s call this a Rorschach—independent teams could create their own concept of the personality and compare Rorschachs. Patterns that are largely duplicative can be considered accurate to the personality, derived through independent means; patterns that are mismatched can be considered inaccurate in some way. 2. Interviews and interactions with others who knew this personality in life, perhaps themselves artificially generated to superb accuracy. Perhaps even MT-A and MT-B could have a chat and see if the other seems "off" somehow. Seems like, starting with biographical data to begin with, you might fairly accurately simulate a personality through reiterative process.

fafromnice wrote:

ethic could be really different in a Jovian Habitat and Anarchist one, in this habitats we could find two different MT who would be utterly different and be perceive to be a "true one"

And yet I still think you could brand a model as "Accurate to .999999%®" and provide the methodology that comports to standard modeling. This is what we can do with public opinion polls, certainly subject to partisan hackery, and the better the methodology the less it is perceived as hackery. Some polls are understood to be more celebrated than others. Some hypercorps would have investment and branding in the accuracy of their generated personas, assuming there's a market, the Swiss Watchmaker of Persona AGIs. Iterative Model #3: Artificially and independently generate a current living ego and compare to the original as control. If highly accurate, generative model is sound.

fafromnice wrote:

Yes I presume, in that case a "real one" will only be the "one" in the eye of the population and what it think about this particularly MT, no ?

True. And in the accuracy you have created value in whatever it was you were attempting to extract from the generative personality, whether genius or clownishness. We've used political people as an example, but I think AGI comedians might be more in demand. Interesting, we're pushing at the boundaries of what is facsimile and what is genuine, the very heart of the apprehension about sleeving and farcasting, and the EP milieu confidence at accurate duplication of "a soul."

Heh. IMEP, I’ve introduced a very competent AGI named John Glenn who seems to know a vaguely uncomfortable amount of detail about the Factors, with whom he has apparently spent an unaccountably large amount of time. He seems to be as much an authority on them as anyone, and they—strangely, mysteriously—about him. Some Persona AGIs I perceive would, once generated, prefer to be deleted or stored. Einstein-99%, I imagine, would chafe at learning God really does play at dice; and I’m not convinced his particular genius can be modeled and he would know that (or suspect that) and despair. Certain personas like Kurzweil-99% or Hawking-99% I imagine would revel in it. I suppose, ethically, any infolife that despaired of their condition and desired instead to be stored should be granted that.

fafromnice wrote:

and some interesting question will appear. Does I have the right to be call Einstein ? am i Einstein ?

This is at the heart of other existential questions of identity and self EP raises about sleeving and farcasting and forking, etc (and why I love it so). I hadn't really thought about this before, but it really makes AGI an interesting control group for other assumptions about authenticity and facsimile. If you could scratch-build an ego ex nihilo with some assurance of accuracy and comparatives, it really lends assurance of authenticity when you have an ego that can be direct copied—"if we're this good with hacks, imagine what we can do with the real item..."

Quote:

(and I want to know what is all the fuzz with your John Glenn-AGI :D )

That would be telling ;-)

I should add here that IMEP I've limited these sorts of high-accuracy Persona AGIs to have required some sort of extant brain in some sort of condition that can be scanned and "improved" through a (re)generative process. Einstein's brain, for example, is in a pickle jar on a shelf somewhere... and I make the assumption that its contents, while degraded and corrupt, might be recovered to some accuracy. There are a number of corpsicles whose brains have been flash frozen, and I imagine these might be the easiest to recover from corruption; and I imagine these might be even be the test models for more sophisticated sorts of recoveries. The oldest of these cryonic subjects was born in the 1890s, and it is intriguing to think some transhuman minds in EP might have been around for a very long time. A lot of these folks counted among the earliest champions of transhumanism and technological immortality, and I think EP should tribute them by imagining them still around in one form or another, but that's just MO. I've fancied some folks like Josef Mengele would be rather fussily inclined to the preservation of their brains, and I imagine the nascent Project Ozma was sneaking brain material from certain luminaries at least as far back as the Kennedy era (whatever happened to his brain after Dallas, anyway?). There was a certain Austrian scientist, a baron, who experimented with brain matter animated by electricity. He was born in 1770 and his brain was frozen, and therefore preserved, in the Arctic. I imagine his might be the oldest transhuman mind operating on a server somewhere. Happy Halloween!

I like the idea that there are some mental entities lurking around that are just as terrifying as the TITANS, and maybe even a bit more malevolent. Anders Sandberg has posted ideas along these lines—minds forked and merged and reforked and remerged again and again until they make the Ultimates seem like conservative pussycats in the comparison. Scary!

I think it's 6x for egoes running on a meatbrain, 60x for a pure infomorph, regardless of whether it's an AGI or a transhuman ego in an infomorph? Have to double-check.

kindalas wrote:

And I'll point out that Mental Speed even thought it only gives two extra complex actions it allows for reading at 10x the normal rate so even in a meat box speed can be achieved.

That's a good point. A really good point. In fact, is 60x acceleration simply having Speed 4 and Mental Speed? Because that would answer a lot of questions. The way I picture growing an AGI is this: First you start of with an AGI "Newborn completely devoid of all knowledge and skills beyond the very basics required for it to remain coherent. This Newborn is then imparted with those skills and aptitudes it will require for it's purpose using Psychosurgery, which can be assumed to be permanent as they will occur under ideal conditions (Rules Wise, we can assume the equivalent of a Slave Eidolon on a Lockbox modified so the inhabitant has Malleable Mind level 2). Installing software upgrades at this point would also make sense, and also provides an interesting thematic for AGIs not wanting to sleeve into Biological brains, because not all of their “mind” comes with them. After this, the AGI will be placed in multiple simulspace environments conducive to the desired skillset; one set to be an astronomer may be “sleeved” into a ship lost in space, and must use it's knowledge to find it's way home, whilst a combat AGI would be placed in various combat arenas. Less scrupulous programmers may use evolutionary forking to increase effectiveness by making multiple forks compete in these environments and only retaining/reintegrating the most successful, but those for those who consider AGIs and Forks people this is essentially mass murder. Even if you can't define an AGI's personality, giving it the tools and abilities to excel in the desired field and a conducive environment means you can guide it along the desired path quite effectively, Back to the OT, you could edit an AGI to think it were a specific person by psychosurgically altering their personality and implanting artificial memories. It's just seriously messed up.