Reign of Steel timeline in 4e?

[Starship]

I was reading about the Soviet Lunokhod programme. They put a robot rover on the moon in 1970. To make the Steel timeline as work as printed what if manned space flight experienced significant setbacks but unmanned was wildly successful?

The POD would probably be 12 April 1961. Yuri Gagarin does not survive reaching orbit. I'm not sure if Kennedy would still commit to go to the moon without that. Even if he did the Mercury and Gemini programs fail in this timeline. America develops their own robots to send to the moon. Robot and computer technology get a huge boost thanks to the space program. In the 1980s robot tanks see production. People start to think about fighting wars by proxy entirely with robots. Computers to run the war machines get advanced. In the 1990s technology has advanced enough that manned space flight is attempted again. This time it works and humans move into already constructed facilities on the moon. 2010 sees the birth of AI and the Robot Revolt.

Since biotechnology wasn't developed as much in this timeline the plagues could be naturally mutated like the Bird Flu. After the uprising the Zoneminds start a crash course to engineer deadly diseases. Or they might skip that step and try to go straight to nanoburn.

What do you think?

[Jerron]

A few comments, but this is an interesting diversion. It sounds like it would work, although have relatively little impact if you were playing after it happened.

1) Kennedy definitely made the 'man on the moon' challenge in response to Russia's success. Failure would likely result in no challenge to get on the moon 'by the end of the decade', although it may still have been a high priority mission- just not the top one.

2) 'Robotic' here is a bit misleading. Remember, Asimov had practically just invented the term. The current definition means it can receive input, process it, and initiate output. That's not much, all it needs is to detect light and deploy the solar panels to be a robot. The definition they're using is unclear, and the capabilities are unclear. It was likely most of its functions were remote controlled, not robotic. However, this in no way detracts from your divergence- it was as sophisticated, for its time, as any piece of hardware.

3) The main difference, it appears, is that you would have most of the wars in the last decade fought by unmanned 'drones', eventually leading into robotic units. See above about robotics. If this were true, there would likely be tried and tested combat units all about. Poor countries, and insurgents, religious types, and guerrillas, would not have any, and be using conventional infantry. The expertise for disabling these units would be tried and true, as well. It would lead to those countries / areas being very hard for the robots to conquer. There would be no existing units to turn on them, as well as when they do arrive, the populace being skilled at disabling them- and knowing they had to. I see all of these areas that are likely to be troublesome, being wiped out by some other method such as nukes.

However, one of the mainstays of technology being oil, I can see a gulf war starring the robot units being fought. Arabic guerrillas cannot be wiped out without destroying the oil, and all of those advantages are still in place for them. Low technology here is actually a bonus for them. As long as they're sitting on top of something the robots want, anyhow.

[oldgringo2001]

My advice is just to ignore the "nothing later than the Homeline present" rule. It's already sort of canon; there is a crossover to Infinite Worlds in Traveller Intersteller Wars. Steel's timeline could be moved even further into the future, allowing for more room to set up your divergent pre-history.

Although I lost my copy, there is a timeline called "Space Nixon" which was published in a Pyramid article back when it was on paper. The premise was that Nixon did not cut back on the US Space Program, so the space race continued, and we had lunar colonies by the end of the last century. This sounds somewhat similar to your proposed line.

If you have Cyberworld, you might consider using its timeline as a base from which to launch the AI revolt. RoS does suggest that the pre-revolt world might have had cyberpunk elements. And if you have Cthulhupunk (which uses the Cyberworld background), you can throw Things Man Was Not Meant To Know into your campaign to make it even more "interesting" for your players...

[Johnny1A.2]

Originally Posted by Starship

I was reading about the Soviet Lunokhod programme. They put a robot rover on the moon in 1970. To make the Steel timeline as work as printed what if manned space flight experienced significant setbacks but unmanned was wildly successful?

The POD would probably be 12 April 1961. Yuri Gagarin does not survive reaching orbit. I'm not sure if Kennedy would still commit to go to the moon without that. Even if he did the Mercury and Gemini programs fail in this timeline. America develops their own robots to send to the moon. Robot and computer technology get a huge boost thanks to the space program. In the 1980s robot tanks see production. People start to think about fighting wars by proxy entirely with robots. Computers to run the war machines get advanced. In the 1990s technology has advanced enough that manned space flight is attempted again. This time it works and humans move into already constructed facilities on the moon. 2010 sees the birth of AI and the Robot Revolt.

Since biotechnology wasn't developed as much in this timeline the plagues could be naturally mutated like the Bird Flu. After the uprising the Zoneminds start a crash course to engineer deadly diseases. Or they might skip that step and try to go straight to nanoburn.

What do you think?

It doesn't quite work...or rather, it sort of misses the issue.

The problem is that you might get faster computers that way, more reliable computers...that's not the same thing as 'problem-solving neural nets'. Today's computers in real life are many orders of magnitude faster and more reliable than ENIAC or its ilk...but they aren't any smarter. Clever programming tricks sometimes give the illusion of such, but that's all it is, as Jaron Lanier likes to point out with sharp accuracy.

(To get an idea of how fast real-life computer tech outstripped projections, look at the old 3e descriptions of the AIs in the original ROS. A typical AI as described there had a mass memory system good for 1000 terabytes, hardened. This one-petabyte memory unit massed 50,000 lbs, occupied 1000 cubic feet, and would have cost ten million dollars under the rules of that period of GURPS 3e. An Overseer might have 10,000 gigs mass storage, a Centurion 5000 gigs.

Today in 2010, you can buy 500 gigabyte hard drives off-the-shelf for about 100 dollars. A petabyte of hardened mass memory would cost only a few hundred thousand dollar or less (depending on the hardening, in fact my calculation assumes using off-shelf gear, you could beat that cost using larger disks), and occupy maybe a few tens of cubic feet. You could probably put a Centurion's mass storage into a cubic foot, hardening and all, at a very modest cost.

Another example: I have a 500 gig portable USB-powered drive on my desk beside my PC. It weights maybe six ounces, give or take an ounce or two. According to GURPS 3e computer rules from the 90s, that should make it a memory unit from TL11, which would make it TL10 in 4e, i.e. age of star flight. Oops. :grin:)

What's needed to make Overmind (or any self-willed AI) is some kind of breakthrough, a whole new approach, rather than just faster and faster binary calculators. So making the time-line work requires more than just fast computer advancement, computer advancement is already going way way faster than anticipated.

[Darekun]

Originally Posted by Johnny1A.2

What's needed to make Overmind (or any self-willed AI) is some kind of breakthrough, a whole new approach, rather than just faster and faster binary calculators.

That sounds like making the assumption that human intelligence is trans-Turing, which is still merely the Quantum Thinkers switch on UT23 or something like it. Tests on small numbers of neurons suggest even our most powerful supercomputers are still a few orders of magnitude short of the human brain, but that there is nothing trans-Turing about them. And Reign Of Steel doesn't seem to make that assumption, so it would be throwing in a roadblock that wasn't already there, and is only a hypothesis in reality.

Or if you simply mean that massively-parallel processors are different from serial processors, that's certainly true, but "faster computers" isn't limited to serial, parallel, or massively-parallel. We've stuck with serial processors for quite some time because it's handy and backwards-compatible, but we've run into limits, and are moving to parallel processors. Certainly, if there were a market, we could produce massively-parallel processors more like the brain — to a certain extent, we do, it's just they'd come from nVidia rather than Intel, and optimization for graphics is away from branching.

[Johnny1A.2]

Originally Posted by Darekun

That sounds like making the assumption that human intelligence is trans-Turing, which is still merely the Quantum Thinkers switch on UT23 or something like it. Tests on small numbers of neurons suggest even our most powerful supercomputers are still a few orders of magnitude short of the human brain, but that there is nothing trans-Turing about them. And Reign Of Steel doesn't seem to make that assumption, so it would be throwing in a roadblock that wasn't already there, and is only a hypothesis in reality.

ROS makes the assumption (in the original time line anyway) of a major computing breakthrough that enables the megacomps to 'rewrite their own operating systems'. This is something that's not even on the theoretical horizon for conventional computing.

Assessing whether consciousness and self-will are explicable in terms of Turing's analysis or not are meaningless, we don't know enough about the issue to even speculate usefully. Even estimates of the capacity of the brain in raw processing terms, relative to any given machine, are mostly guesswork.
Our understanding of how the brain works is so crude and limited that we can't even make a reasonable estimation of brain capacity that we can use for much.

Whether it's an issue of hardware, software, or anything else, the emergence of a self-willed AI in a time-frame such as ROS requires something other than just 'faster hardware'.

[Johnny1A.2]

Originally Posted by Darekun

Certainly, if there were a market, we could produce massively-parallel processors more like the brain — to a certain extent, we do, it's just they'd come from nVidia rather than Intel, and optimization for graphics is away from branching.

No processor currently built or that we know how to build is very much like the brain.

[Langy]

Originally Posted by Johnny1A.2

ROS makes the assumption (in the original time line anyway) of a major computing breakthrough that enables the megacomps to 'rewrite their own operating systems'. This is something that's not even on the theoretical horizon for conventional computing.

Actually, it is. It's called self-modifying software. It simply requires the program to have access to its own source code, a compiler for that source code, and an algorithm for modifying that source code. The difficult part is the algorithm, not the ability to rewrite the code in general. I'm pretty sure there have been research projects on self-modifying code out there, especially ones doing 'evolution'/'survival of the fittest' type projects with multiple relatively simple agent programs. If I remember right, that's how at least one robot was 'taught' how to walk.

[Johnny1A.2]

Originally Posted by Langy

Actually, it is. It's called self-modifying software. It simply requires the program to have access to its own source code, a compiler for that source code, and an algorithm for modifying that source code. The difficult part is the algorithm, not the ability to rewrite the code in general. I'm pretty sure there have been research projects on self-modifying code out there, especially ones doing 'evolution'/'survival of the fittest' type projects with multiple relatively simple agent programs. If I remember right, that's how at least one robot was 'taught' how to walk.

Which doesn't modify my statement. The sort of self-rewriting the megacomps of ROS are supposed to be doing is to that as jetliners are to paper airplanes.

[Darekun]

Originally Posted by Johnny1A.2

ROS makes the assumption (in the original time line anyway) of a major computing breakthrough that enables the megacomps to 'rewrite their own operating systems'. This is something that's not even on the theoretical horizon for conventional computing.

…The computer I'm typing this on has everything needed for that but the intelligence and the skill. It doesn't understand enough to usefully rewrite its own operating system, but if a virus somehow got enough permissions and was pre-programmed with what to do, it could rewrite the operating system. I could certainly write a program to randomly rewrite the operating system until it broke.

The line before that in the timeline is the invention of a neural net with the intelligence. Teach it Computer Programming skill and it can rewrite the operating system.

So no, that's not a separate breakthrough at all.

Originally Posted by Johnny1A.2

Assessing whether consciousness and self-will are explicable in terms of Turing's analysis or not are meaningless, […] Whether it's an issue of hardware, software, or anything else, the emergence of a self-willed AI in a time-frame such as ROS requires something other than just 'faster hardware'.

The second puts lie to the first. If human intelligence is Turing-level, then it can be emulated by any other Turing-level processor, with some speed reduction, so faster hardware is all that's needed to get human-level intelligence in realtime. If human intelligence is trans-Turing, then it would be true that something else is required. By asserting that something else is required, you are asserting that human intelligence is trans-Turing. Nothing more, nothing less.

Originally Posted by Johnny1A.2

No processor currently built or that we know how to build is very much like the brain.

Your "very much" is covered by my "to a certain extent". The brain is a massively-parallel branch-oriented processor, the CPU in this computer is some RISC thing emulating a serial branch-oriented processor, the GPU in this computer is a massively-parallel gradient-oriented processor. We could make massively-parallel branch-oriented processors, even make them arbitrarily similar to the brain, but the only market would be AI researchers, and they just take the emulation hit instead. The speed cost of emulation to go between serial and massively-parallel processors is high, but the market is that small.