[Spaceships] Sneaking up on a Megastructure

[AlexanderHowl]

Define population. A SM+53 MB would have a Complexity 32 computer, allowing it to run a single IQ 54 SAI, 1,000 IQ 48 SAIs, 1 million IQ 42 SAIs, 1 billion IQ 36 SAIs, 1 trillion IQ 30 SAIs, 1 quadrillion IQ 24 SAIs, and 1 quintillion IQ 18 SAIs. It would also likely have 200 septillion cabins for biological creatures (which would only be one SM+53 habitat component), allowing the MB to easily support 1 septillion biological sapients.

The MB would likely use highly compressed encrypted quantum laser comms for efficiency and security of communications, so it would be likely unlikely that the probe would intercept any useful communications. Anyway, it would require a Complexity 14 computer to break TL12 basic encryption, and the quantum laser comms would allow the MB to automatically detect a listener. Radio communications would be highly unlikely, as radio communications have decreased from TL7 to TL8, meaning that a TL12 society would probably not use any radio communications that could be detected by a probe (maybe the equivalent of WIFI, but it would not penetrate the shell of the MB).

[DataPacRat]

Originally Posted by AlexanderHowl

Define population.

Well, that's the trick, isn't it?

A SM+53 MB would have a Complexity 32 computer, allowing it to run a single IQ 54 SAI, 1,000 IQ 48 SAIs, 1 million IQ 42 SAIs, 1 billion IQ 36 SAIs, 1 trillion IQ 30 SAIs, 1 quadrillion IQ 24 SAIs, and 1 quintillion IQ 18 SAIs. It would also likely have 200 septillion cabins for biological creatures (which would only be one SM+53 habitat component), allowing the MB to easily support 1 septillion biological sapients.

Well, ignoring the bio-critters as likely being an inefficient use of mass, a quintillion would add something like +27 to SS5's detection of the existence of signals; and since at 25 light-years there's only a -7 range penalty, that sounds like odds of detection of 'yes'.

The MB would likely use highly compressed encrypted quantum laser comms for efficiency and security of communications, so it would be likely unlikely that the probe would intercept any useful communications.

UT's lasers have a fairly limited range compared to the size of a MB, even at TL12, merely 50k miles; while a black hole gravity-ripple comm has a range of a full parsec. (Which, of course, would be impossible for anyone without a half-ton grav-scanner to pick up at all.)

Anyway, it would require a Complexity 14 computer to break TL12 basic encryption

There's a difference between decoding comm protocols, and decrypting the content - the former is what's described in SS5, the latter in UT. And the latter says smaller computers can do the job, they just take longer; a TL10 Tiny Quantum Computer is $500 and 0.1 lbs, Complexity 5, and effective Complexity 10 for decryption - which means it takes 81 hours instead of 1 per decryption attempt. (And 729 hours per try if working against TL12 secure encryption.)

, and the quantum laser comms would allow the MB to automatically detect a listener. Radio communications would be highly unlikely, as radio communications have decreased from TL7 to TL8, meaning that a TL12 society would probably not use any radio communications that could be detected by a probe (maybe the equivalent of WIFI, but it would not penetrate the shell of the MB).

Bandwidth is bandwidth; there are only so many bits-per-second you can shove through any particular channel, and it seems unlikely to me for a highly-digital culture to want to hobble itself just because some frequencies have leaky sidelobes.

[AlexanderHowl]

If there is any sort of competition between the upper intelligences, they will likely encrypt their messages, probably at a level that would be impossible for lesser intelligences to decipher. For example, an IQ 42 intelligence running on a Complexity 26 computer would probably use encryption that requires a Complexity 26 program to crack. Good luck cracking that with a Complexity 6 quantum computer (it would take over 1,600 years).

[Anthony]

In practice, you will be able to detect a stellar megastructure at any range where the star itself could be detected, as it has stellar level energy output, with a very strange spectrum.

[DataPacRat]

Originally Posted by AlexanderHowl

For example, an IQ 42 intelligence running on a Complexity 26 computer

Speaking of which; given that we're already stretching beyond the limits of UltraTech by talking about computers bigger than megaframes, and once an AI's IQ gets high enough to have a skill-level of "yes" even with penalties, I'm thinking that it's likely safe to assume that 3 or 6 or 9 or so levels of the computer's complexity are dedicated to speeding up the AI rather than adding directly to its IQ. Eg, being able to take 240 hours of extra time in 0.864 seconds. (With some variability depending on need, and the AI's version of wealth to pay for the extra CPU time.)

[AlexanderHowl]

That would be in the form of assigning projects to LAI and NAI fragments of itself. For example, an IQ 42 SAI can run 1000 IQ 38 LAI fragments or 1 million IQ 34 fragments using the same processing power. If the task only requires IQ 38, it will assign a LAI fragment. If the task only requires IQ 34, it will assign a NAI fragment. The same would apply to any IQ 30 SAI, though it would have IQ 26 LAI fragments and IQ 22 NAI fragments, and the central IQ 54 intelligence, though it would have IQ 50 LAI fragments and IQ 46 NAI fragments.

[DataPacRat]

Originally Posted by AlexanderHowl

That would be in the form of assigning projects to LAI and NAI fragments of itself. For example, an IQ 42 SAI can run 1000 IQ 38 LAI fragments or 1 million IQ 34 fragments using the same processing power. If the task only requires IQ 38, it will assign a LAI fragment. If the task only requires IQ 34, it will assign a NAI fragment. The same would apply to any IQ 30 SAI, though it would have IQ 26 LAI fragments and IQ 22 NAI fragments, and the central IQ 54 intelligence, though it would have IQ 50 LAI fragments and IQ 46 NAI fragments.

I don't think that works out quite the same as actual subjective sped-up time. In a non-cinematic universe, you can only reduce the time to 1/10th normal with a -9 haste penalty; that's the same time-reduction as spending 1 level of complexity on a x10 mental speed-up, and the latter doesn't suffer the skill-penalty.


On a related note, there's an item from 3e's Ultra-Tech 2 that I think is worth checking. To link up a pile of identical megaframes into a machine that's effectively higher-complexity, then take the total number of computers, square it, and multiply the result by $50 and 0.1 lbs for all the additional connections and suchlike hardware. So, to meld a billion mainframes into a machine with +9 complexity, you'd need an extra $50Qn and 100Qd lbs. (Which would be on top of the $5Qn, 40T lbs for a billion TL12 C15 Genius Megaframes.)

It's late, so I'm probably going to wait until tomorrow to check how much of this gear could be stuffed into a dozen or so SM+53 cargo bays; it might put a somewhat lower cap on how high the MB's AIs' IQs can reach, and thus their skill levels, and thus their detection-ranges.


Edit: Oh, fine; let's go for whatever power of ten leads to less mass than the Solar System's:
100,000,000,000,000 computers:
TL12 Genius Megaframes: C15, $5B, 40k lbs ea
1e14 of them: $5e23, 4e18 lbs
Networking Gear for 1e14 computers:
$5e29, 1e27 lbs
Total Complexity: 29

[AlexanderHowl]

You do not need to link up multiple computers, the computers are already linked up in the control room. A TL 12 Complexity 32 computer would mass around 2×10^20 tons and cost around $10^26, which is not actually that much compared to the cost of the control room. Of course, the MB would have likely evolved over thousands of years, so the cost would have been spread out over that time, and the economic potential of the MB would be phenomenal. For example, the cost of the energy production alone would be $2×10^25/year at TL8 (assuming 20% efficiency and $30/MW-h (e) wholesale).

[DataPacRat]

I think I'm going to conclude that GURPS has better ways to model very-superhumanly-intelligent MB-living AIs than simply cranking up the IQ and skill levels. I'm thinking of tweaking the rough 'Path to Victory' build I mentioned in the other thread, recreating it to better model 'hyper-advanced digital modelling'. Eg, tweaking Precognition so that it's closer to the Digital version of Oracle, relying on extrapolating from existing sensory data (possibly even have it rely on the Hypersensory version of Psychometry, ala "Accessibility: Only While Using Psychometry: -10%" from Power-Ups 8 p5). Maybe throw together a variant of the "Logos" or "Noumena" powers from Powers: The Weird, to reflect their basis, ala that book's description of GOFAI (Good Old Fashioned AI) Style.

(One reason I'm leaning toward stepping away from high-IQ AIs, is that with THS-style AIs, you can crank up IQ arbitrarily high just by paying more cash. Something like an IQ of 10,000 for around $6B, if I remember the pricing right. And throw in another $4M for 40 CP in each of 1,000 skills, just to avoid the 'untrained' penalties. Which puts things so far outside of GURPS's assumptions around 3d6 rolls that, well, we might as well not bother with the dice. Which seems to map closer to 'Cosmic: No Die Roll Required' than bothering to quantify the exact numbers of an IQ level of yes.)


I'd still like to base some numbers in Spaceships and Ultra-Tech if possible, simply to more easily interface with the rest of GURPS. I'm not sure if Gadgeteering is a good fit, since it's been a while since I refreshed my memories on those rules, but I'll take a peek; and if that doesn't work, I just might try eyeballing some estimates of program-complexity versus accuracy-of-Precognition, or the like.

And, of course, I'd appreciate any suggestions anyone has on this approach. :)