Initially, this is going to operate with rapid player turn-around; if you’re in SkyNet, you may be selected to be a reprog, or dragged into reprog roleplay, and turned around on a high risk mission; these are not characters - they are tools. The present administration is still playing around with the idea of reprog characters, but before the introduction of the 800 series on the server, that won’t be happening. If you get reprogrammed earlier, be prepared not only to be an expendable asset, but to be replaced with certainty; don’t get attached to the position. Reprogrammers, the administration, and SkyNET command are the people to approach if you're both in the IC position for this to happen, and interested in it on an OOC level (they will ask for an auth thread, so you should just make a request here). All above conditions met, or if you're selected arbitrarily, and the canon above is understood, one of the authorized techies will sort it out IC. Spoiler: RULES 0-3 Reprogs follow four rules: Currently, bots are to be programmed to obey four very important rules: 0. A robot may not injure a human being or, through inaction, allow a human being to come to harm. I. Except where such directives would conflict with the Zeroth Law, a robot must obey Primary Directives. II. Except where such orders would conflict with the Zeroth or First Law, a robot must obey orders given it by human beings. Reprogs can only be commanded by any member of Charlie Squad, Sgt+ of other squads, and Tech Comm Higher Command. III. Except where such orders would conflict with the Zeroth, First, or Second Law, a robot must protect its own existence. Reprogs can only be commanded by any member of Charlie Squad, Sgt+ of other squads, and Tech Comm Higher Command. Spoiler: PRIMARY DIRECTIVES Functionality does not translate perfectly through each series drone. Smarter/newer drones are more capable of filling in blanks their older and dumber counterparts cannot. When in doubt, double and triple check. - An order is invalid if not delivered by members of Tech Comm who are either members of Charlie Squad or a SERGEANT in any squad. - Life-Saving Protocol : identify wounded Tech-Comm soldiers and bring them to either a corpsman or a designated medical bay. Further orders may be delivered. - Aggro Protocol : WHEN ORDERED TO DO SO, engage as many hostile targets as possible. CALLSIGN SERIES OOC NAME SEKHMET 400 Prezy BOLT 500 Wizard TALUS 600 ValueSensei MEGATRON 600 Chicken McDendi NAPALM 600 Wizard KITT 600 BoB CREED 600 A Blaze of Glory ATLAS 700 Kenny STATIC 704 Hardhat FREYJA 800 Deathgarand IRIS 800 Commander Potato VALKYRIE 800 Disorder ARES 800 Alfa TYCHE 800 Valerstein Sasha 850 Enraged NAME SERIES PLAYER Technical Communications, Charlie Labs here with another status update. This time? Project Titan! Project Titan is the name for the research we over at Charlie have been conducting into Skynet Drone hardware, firmware, and software, with the desired end result of taking what we’ve learned and turning it to both swords and ploughshares! Quite a number of you have made significant contributions to this effort, mainly through turning over undamaged components, and even full bot endoskeletons, to our team, and we’d like to take a moment, prior to getting into the nitty gritty, to thank you for your efforts. At this point in time, we can confidently state that: We have the capability to repair bots, up to the 700 series, to working order. We have a near complete architecture understanding of how bots up to the 700 series operate. We have the capability to dynamically interface with and reprogram drone directives, up to the 700 series, and can do so repeatedly and reliably in a process that won’t take more than an hour within an acceptable margin of error. Some technical background (feel free to skip this section if you’d just like to see where you fit in, and how these things should begin to aid our efforts): SkyNet Drones are, at their core, computers. They have storage, transitory memory, input and output devices, connective microcircuitry (a motherboard), and a central processing unit. While the majority of these devices are not particularly remarkable and generally adhere to the Moorian standards we’d expect to have by this time in technological development (~10^9 transistors/mm^3), the real marvel is the central processing unit, which appears to utilize quantum superpositions in its calculations, giving it previously impossible real time computation abilities of NP-hard problems. Rather than either being a 0 or a 1, these devices’ base-unit (old literature suggests referring to them as qbits) are a range of probabilities represented by: where meaning that every qbit in a chip represents 2^N bits: 2 qbits == 4 bits, 3qb == 8b, etc. Put simply? These machines are not constrained by the binary-driven proceduralism set up by the fundamental architecture of classical computing systems, and in fact experience an exponential growth in computational ability in line with linear hardware efficiency breakthroughs (and that is a generous underestimation of our opposition). The silicon supporting these central processing units are more akin to life support in classically biological organisms: necessary, but fundamentally simplistic. The symbiotic nature of these binary/quantum computation systems allows the machines to perform real-time artificial neural networking - or, in other words, the ability to dynamically alter behavior non-programmatically, in a way that emulates organic neuroplasticity. The development of this technology can be observed between series. Earlier series drones, such as the 1 or 500 series, are capable of learning how to navigate through new terrain, such as sand, mud, or debris, without preconfiguration and with only a few minutes of trial and error, while later series, such as the 700, display object permanence and even minor predictive capabilities. According to our reverse engineering, it appears that there is a limiter switch built in to these processing systems. The vast majority of these systems use their neural networks in a read-only mode and do not incorporate new operant conditioning into their subroutines. This appears to be a part of deliberate pre-training by skynet to limit aberrant behavior. At this time, for the same purpose, we will be keeping this mechanism off as well. With the combined exponential growth of silicon density and quantum amplification, it would not be unexpected for the upcoming seriess of terminators to display intelligence rivaling, or even surpassing, that of our own. Indeed in as little as a few years SkyNet is likely to possess the capability to model the human connectome in its entirety… yet by that point our evolutionarily derived equivalent processing power of ~40 petaflops will likely be deemed architecturally inefficient in favor of a more engineered solution. Fig I. A T700 CPU, top of the line. In some sense, these CPUs are black boxes - the exact behavior and decision making of each Drone, individually, varies depending on the experiences and training it has had outside of the original operant conditioning hardcoded into their neural networks. They are provided sensory input, hard rules, and directives by their supporting silicon architecture, which are then processed and returned to output devices for actualization. These behavioral distinctions become more obvious the later the series number - the more complex the “black box”’s internals are. Direct editing of active neural networks without catastrophic failure is likely outside of even SkyNet’s capability at this juncture - which leaves us to tinker with the silicon. Thankfully, silicon behaves as expected. A one is always a one and a zero is always a zero. The biggest challenge of this project was creating interfacing tools to scrape data off of… which is where you come in. The process that has been developed by Charlie is as follows: 1. Acquire drone hardware 2. Perform any necessary repair 3. Resect main chipset, hook into DecompSaddle (DecompSaddle developed to interface between most commonly found skynet interfaces and standard SCSI port) 4. Run zHak (zHak [“Zack”] dynamically changes SkyNOS environment variables and replaces the standard drone bootloader to disable hashsum checking and reflashing; recompile times range from 5 minutes for earlier models, to 45 minutes for 600+ models; upon failure zHak will save logs and throw an error) 5. Reassemble drone barring servo control - hooked into zEndoSandbox interface port 6. Monitor attempted drone servo activation for a minimum of 15 minutes to assure reprogramming process has been completed successfully, testing advised during this timeframe 7. Shut down drone, reconnect servos, give Primary Directives 8. Activate drone, mission ready Fig II. A prototype DecompSaddle So… what can you do to help? Bring us Drones. Don’t bring them “alive” unless they’re remarkably well secured, and don’t bring them shot to utter pieces. If you find a new model floating around out there, document as much as you can before performing perforative field maintenance on it, and bring the reports right to Charlie. Currently, bots are to be programmed to obey four very important rules: 0. A robot may not injure a human being or, through inaction, allow a human being to come to harm. I. Except where such directives would conflict with the Zeroth Law, a robot must obey Primary Directives. II. Except where such orders would conflict with the Zeroth or First Law, a robot must obey orders given it by human beings. Reprogs can only be commanded by any member of Charlie Squad, Sgt+ of other squads, and Tech Comm Higher Command. III. Except where such orders would conflict with the Zeroth, First, or Second Law, a robot must protect its own existence. Primary Directives are issued by Technical Communications command and/or hard-coded by Charlie Labs. However, as chain of command is a nebulous concept, and one we’ve been struggling to work into their programming, but bots with language interpretation capability have haphazardly been programmed to obey verbal commands from Sgt+ in accordance to the Laws (please don’t confuse the bots with multiple conflicting orders - remember that at their core, these things operate with unpredictable, self-taught algorithms… you don’t want to inadvertently become a case study in machine learning). Bots without, such as the 1-500 series, simply operate on Primary Directives. Fig III. ZHak Progress Screen Due to earlier and, to put it politely, more haphazard attempts at drone reprogramming, there is likely to be some concern among TechComm personnel about having active death-machines walking among their ranks. The aim of this document is in no small part to reassure you that while the mechanitions of synthetic non-human consciousness remain inscrutable to us still, that as long as 1==1 and 0==0, and all the steps outlined above are taken by personnel trained to perform them, that they are no more dangerous than your standard issue weapon, and will serve to sow chaos only against our enemies, if directed appropriately. ___________________________________________________________________________________ T-800 Spot Bulletin: After nearly a month of study, the first captured T-800 CPU has been successfully reverse engineered. Unlike prior series, the 800's computational components are all unified on a singular chip. There is no external RAM, ROM, storage, or I/O subprocessors. Every necessary component resides on the main chip housing the neural net processor. This appears to be done for two reasons: One: because the T-800's computational archetecture is all centralized, its combat resiliency is significantly increased. Beyond the power supply and the main chip, the entire robot is nothing more than a "dumb" exoskeleton. As such, there are far fewer effective places to direct fire effectively. Rather than the entire head being a distributed computing system of disparate components, there is simply a single 4x5cm square buried within layers upon layers of titanium alloy. Two: it is much harder to reverse engineer. The chip is housed in a deoxygenated environment in the robot's head and coated with white phosphorous. When the head is opened, either to remove the chip, or due to combat damage, the phosphor will begin to oxidase. While there are rumors that this will cause the chips to explode in flame. This is not likely to happen outside of very specific environmental conditions. Rather, the phosphorous corrodes the chip destructively in a matter of minutes, rendering it fully destroyed and incapable of being reprogrammed. The entirety of the 800's operating system is run time XOR encrypted. It is believed that this and the phosphorous coating on the chip were done as countermeasures after SkyNet experienced significant battlefield losses from existing Technical Communications reprogrammed robots. The other discovery made during this process of reverse engineering was done via non-transient electrostatic field analysis augmented by processing power from daisy-chained T-700 neural net processors. We've begun to crack the evolutionary derived algorithms that take form when the robots are in operation: that is, we've started down the path of rudimentary live neural net disruption and modification, and can provide weighted biases while training them. As such, the steps for reprogramming a T-800 differ from previous models: Assure that the robot is powered off. When this is confirmed, the head must be detached and immersed in mineral oil. Open and remove the CPU while fully immersed. Keeping the chip under the mineral oil, use a toothbrush or similar non-abrasive synthetic brush to entirely remove the phosphorous coating from the chip. When the phosphorous is completely removed, remove the chip from the mineral oil and clean it. Utilize the updated DecompSaddle Mk. II to interface the CPU into DreamLand, a fully sandboxed environment powered by six daisy-chained T-700 CPUs. DreamLand simulates a bot's servos and provides a safe environment to allow a non-reprogrammed CPU to function with complete autonomy. Note, there is no backward compatibility between models of DecompSaddle. Power the CPU on. Log onto the DreamLand console and run a XOR string search on the chip's binary architecture RAM for the phrase "SKYNET_CORP_MDL_". The 128 bit return will be the machine's decryption key. Once the decryption key has been recovered, power on zHAK (a copy should be provided on the machine running DreamLand, and run the automated process to reallocate targeting parameters on the silicon. Because of the sophistication of the 800 CPU, and the new ability of the machine's neural net to modify its supporting silicon, the neural net will require training. In order to do this, DreamLand possesses a command line function "electric_sheep" or "es". es will toggle the machine's "read only mode" to "read/write mode" and artificially accelerate perceived time for the CPU and begin to run it through a Darwinian selection process, whereby it is rewarded for adhering to the base rules outlined in the above document, and punish it for failing to, while providing weighted bias on the neural net. The rules are trained in order, from zero on up, to establish hierarchy. When the machine reaches six nines level of behavioral adherence, it is deemed successfully reprogrammed. "Read/write mode" is switched back to "read only mode" and the CPU is ready to be transplanted back into the endoskeleton. This entire process takes roughly 60 minutes, simulates approximately 50,000 years of real time for the robot, and, depending on the decisions made while being trained, can offer the bot 10^11"lives". Note: NEVER toggle read/write mode on a live bot. This can result in unpredictable behavior. ___________________________________________________________________________________ Love, ZE, from your friendly neighborhood Charlie Labs! Spoiler: MAIN TAKEAWAYS Reprogs follow four rules: Currently, bots are to be programmed to obey four very important rules: 0. A robot may not injure a human being or, through inaction, allow a human being to come to harm. I. Except where such directives would conflict with the Zeroth Law, a robot must obey Primary Directives. II. Except where such orders would conflict with the Zeroth or First Law, a robot must obey orders given it by human beings. Reprogs can only be commanded by any member of Charlie Squad, Sgt+ of other squads, and Tech Comm Higher Command. III. Except where such orders would conflict with the Zeroth, First, or Second Law, a robot must protect its own existence. Reprogs can only be commanded by any member of Charlie Squad, Sgt+ of other squads, and Tech Comm Higher Command. ADDENDUM Drones that are PK'd are killed under the premise their CPU has been destroyed. CPU's cannot be salvaged once destroyed - if a reprog is killed, they are dead, and that's the end of it.