"Eliezer S. Yudkowsky" <sentience@pobox.com> writes:
> I don't think that having access to ELTM (Extended Long-Term Memory, aka
> "The 'Net") constitutes true intelligence enhancement. Likewise for
> pencil and paper or a PalmPilot; likewise high-speed arithmetic;
> likewise chess-playing advice.
>
> If these abilities were part of our minds, if they were integrated with
> everything else, they might provide some true intelligence. Even then,
> I'm disinclined to believe. People with eidetic memories and "lightning
> calculators" are not noticeably transhuman.
I disagree, they are transhuman in some ways. The trick is of course to increase overall effective intelligence and not just a few, rather trivial aspects of it. I would say having a good net access is a more general form of IA than a built-in calculator, and if I could download skills from it it would be even more general. What constitutes real intelligence seems to be a largely semantic question, what really matters is what problems the individual can ask and solve.
> These abilities are all very easy to write into a science-fictional
> character. "Doc" Smith was writing "superintelligent" characters back
> in the 30's. Real intelligence (technical term: "smartness") is defined
> by your inability to write a character with the same abilities. If you
> could predict what a transhuman character would do, you would be transhuman.
What about a transhuman character with partially deterministic aspects:
(sorry for pirating your nice pastisches, I enjoyed them a lot :-)
Baron Hans Nidrach von Pompzidaize sat in his laboratory, looking at experimental test subject X17. "How do you feel?" he inquired, his rolling bass echoing from the laboratory walls.
"Superintelligent, Doc," replied X17, who had once been known as John Smith. "I've only had the Throatwarbler-Mangrove Super-Neural Bypass for sixteen seconds, and I've already learned twenty-seven languages and figured out how to play the piano. I have also realized that you have placed a kind of mental lock on me, making me unable to change the subliminal programming you placed in me during the procedure."
Baron von Pompzidaize frowned, examining several multicolored readouts. "Ach, you noticed it? So what will you do about it?"
"Nothing. Since one of these programs prevents me from changing them."
"And you can't find a solution?" the Baron asked, a faint fint of unease obvious to X17.
"No. If you had programmed me as you originally had planned, it would have been easy for me to manipulate you to remove the lock. Unfortunately you made a mistake making even such manipulation impossible for me to do. In fact, if you tried to remove the lock, I would do everything I could do to prevent it."
The Baron re-read the printouts and looked pleased. "Well, then, do you now feel competent to go destroy the Evil Empire and rescue the Princess? Acting in accordance with the 1930s North-American conception of gentlemanly behavior, of course."
"Sure, Doc," said X17. "It's not like I've got anything better to do."
"Excellent," said the Baron, checking two gauges and a flashing display. "You still have the emotional maturity of a flatworm, like everyone else in this novel. I was afraid your superhuman abilities might give you goals slightly at variance with mine."
After X17 left, the Baron realized that he was the victim of a most subtle and cruel revenge. Because X17 had done something awful to him, without even doing anything. Now he would live in fear that his creation actually *wasn't* bound by the rules, regardless of what the actual state was. He could never tell, and trying to figure it out would only drive him further into paranoia. He couldn't even be sure if X17 had foreseen his every reaction and played with him, or if he really had power over X17 but were a victim of his servant's manipulations. X17 had indeed punished the Baron, in a rational manner as suited his rational creator.
(OK, not a perfect example, but I couldn't resist)
> He stood up, executing the movement with impossible smoothness.
Actually, this is something I have been thinking of. Superintelligen motor programs. They might be quite interesting, perhaps even optimal. Imagine an SI in the kitchen...
-- ----------------------------------------------------------------------- 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