Why might a robot's memory be unreliable after the Apocalypse?

Question

Setting

A nuclear war between two superpowers happens, and nuke after nuke is fired at nearly every country on the globe. The war, which lasts for about 4 hours, kills over 3 billion people, and many more were killed from the resulting violence, pestilence and starvation. During that time, known as the “Great Fall” by future generations, no records of the world immediately after the apocalypse were made, as people were either dead or too worried about survival to make any entries into journals. So, most accounts of the “Great Fall” era were written DECADES or more after the fact. Which means, most modern of those accounts were unreliable.

26th century

It is now the the twenty-sixth century, and historians are finally realizing that those accounts of the Great Fall were unreliable. They found this out because many of the accounts mention the Midwestern Empire being involved in the war, but the Midwestern Empire wasn’t formed until 40 years after the war. Their are also mentions of Kodiakist Churches being destroyed in the fighting, but the Holy Mutant Kodiak wasn’t even born yet. So, the historians search and search to find out what really happened during the war, what year was that it when it happened, and who fired the first shot. This is a major plot point.

The Androids

The androids were mechanical beings created or do tasks like cleaning and fighting before the war, and though many of then died and were destroyed during the war, many did survive, even all the way up to the 26th century. Almost all androids are fitted with a video recording camera, and can, at any point, rewind and show their memories to anyone who asks. Most androids turned them off once the war was finished, which made it easier for them to survive as the Nor has a limited supply of fuel. But, they can still access the memory saved into them. The historians really just have to asks an android computer to see one segment of their memory of the war, and everything would be solved.

But, I don’t want it to be that easy, for the sake of not making my story five pages long, so how could you make it so that the androids' memories are also unreliable?

Answer 1

Selective memory. Androids don't keep the records of everything they see. That would create far too much data with very little benefit. They only keep full video recordings for a limited timespan. Everything older gets deleted, unless the android considers it to be relevant for their work. And politics of a past society are not relevant for androids.

If you check the records of a domestic service android which was operational before the great war, you might gain excellent data about the accumulation of dust, the solving capabilities of different detergents and the distribution patterns of children toys.

But you won't get any data about political events, because there is no reason why the android would need to know that. The android might have overheard some dinner table discussions about how those nuclear disarmament talks are weakening the nation, but then deleted it as irrelevant "humans doing human stuff" data. They might have followed recent politics in case any new law affected their work. But when all the politicians died in nuclear fire, society broke down and postapocalyptic gangs took over, they thought: "well, the guy with the spiky shoulderpads says the old laws don't apply anymore, so I can just delete that archive of congress debates".

But there are still some fragments you could puzzle together from android memories. Like the one android who remembers being ordered to replace 10,000 "Death to Arstotzka!" posters with "Welcome Arstotzkan liberators!" and the other whose job it was to measure radiation levels of stuff coming out of centrifuges.

Answer 2

I'm going to suggest something a little different, quite simply it's because:

Individual accounts don't work

According to your own data:

Only about 600 androids survived, scattered across the continent

And

Almost all androids are fitted with a video recording camera, and can, at any point, rewind and show their memories to anyone who asks.

So, now we have 600 videos. But, the thing is that each of those videos will display something different. Some videos may mention something about retaliating against the American pigs. Other videos may proclaim it a great day in history. The Canadian videos may talk about the most recent trade wars. And so on.

No video may corroborate the others and worse, no one may even understand the terms used. It could even be that all 600 robots were janitors in schools and you have 600 videos of screaming children. Even with photographic proof, only 600 points of view makes it unlikely to ever know what actually happened. Unless one was actually in the military control rooms at the time.

Answer 3

Physical Degredation

Electronic storage media does not have an indefinite life span; the physical material used deteriorates just like anything else, which typically leads to data corruption. Over a span of 500+ years, I'd be greatly shocked to learn that the memories were intact.

CDs, for instance, tend to last about 15 to 20 years if stored properly. A magnetic hard disk (what you see in computers without SSDs) is likewise at risk of failure, judging from how many people have to call tech support when their hard drive fails: you can also get an immediate problem if the writing head makes actual contact with the disk (there are fail-safes to avoid this, such as fall detection sensors to park the head before the disk hits the ground, but that is itself a physical mechanism that can degrade). Flash memory is a good deal more resilient in the face of abuse, but even an SSD is liable to suffer memory read/write problems after five hundred years of regular use; they can only write a finite number of times before the physical memory is damaged too badly to retain information, and they aren't designed to last multiple lifetimes as a general rule.

There's also radiation from your nuclear war, which is most certainly capable of causing corruption of data. Even ignoring that aspect, ordinary cosmic radiation can cause damage over a sufficient period of time; not generally relevant to the normal use case in the modern world where most hardware gets tossed out within ten years, but over hundreds of years (or for anything involving space travel) this becomes a notable hazard. Designing software with space-travel-grade fail-safes to compensate for hardware failures or data corruption is obscenely expensive (and even then is not perfect, because programmers are not perfect) and thus is not about to have been done for your standard androids.

Granted, you can always get a new memory unit and transfer the data over, but capacity (while cheap) is still finite, especially when space is limited (as opposed to giant server rooms); some information will inevitably be prioritized, and older data is likely to be sacrificed first. Besides, in the aftermath of your apocalypse, proper maintenance (or even the capability for it, since you've probably lost most of your hardware production and software knowledge when half the planet's population got killed off) is unlikely until society gets back on its feet, which means the androids will have had to do that prioritization themselves with no means to preserve whatever data was sacrificed. Some of the information your historians are looking for simply won't exist any longer.

Answer 4

It could just be an aspect of the android brain. They have a limited amount of storage, and are running well past their planned life cycle. They may have a series of algorithms similar to human minds that combine similar memories, and throw out the original recorded data. This allows the android brain to have the capacity to create new memories, but retain the "experience" of the collection of previous memories. as "memories" combine with "experiences" over time they become more and more unreliable.

Answer 5

Androids will be reliant on SSDs or other flash-like memory, unless designed to be specifically EMP resistant (soldier androids, for example) in which case they might use exotic if slower 3D optical crystal storage.

Any flash or flash like SSDs that were powered up when the initial nuclear war happened will have been damaged, wiped or mostly corrupted - but mostly just wiped - and many of those will also have had the access and control boards fried in the EMP event (what the androids care about for the most part - though many of them will also avoid areas of ongoing ionizing radiation too) and so any given android, having multiple storage devices (long term, mid term and RAM) may have garbled areas, some retained and accessible memory, and in many cases - none at all.

Basic operating instructions would have been firmware, so those androids who were downcycled for recharge or repair during the attacks will have remained largely undamaged - but most operational androids will have been rendered utterly useless - some of their gross mechnical components would have been scavengable, but no sane android would even try a scavenged board or SSD... far too much risk.

Answer 6

Before the fall the android's data was stored in the cloud. Androids have a local cache of limited size. If the cache is full then the oldest memories are deleted first.

This would result in androids "remembering" the last X hours worth of memories. Busy active androids which create lots of memories might have ten years worth of data in their cache. An android that's spent 6000 years in a cupboard might have memories from before the fall.

Storing data in the cloud has many advantages: easy backup, unlimited storage, easy monitoring by evil-corp. Before the fall the cloud was ubiquitous, androids were unlikely to be disconnected from it for long. A local cache was required because sometimes an android might lose the connection to the cloud temporarily, perhaps the enemy starts jamming the wifi network. The developer who coded the cache size limit never thought it would be reached. They did it for completeness.

Answer 7

Answers No. 1 and 2 will most likely not work, because it is highly unlikely that among 600 androids all lost the memory of the war just per coincidence. You could it make work, if just one or two of the androids had surviving memories at all (which by the way) is the most likely case anyway, considering the devestating effects an EMP and long time degredation has on electronic storage - even optical ones).

1. Damage through nuclear EMP

The electronic storage unit of the androids could be shielded against EMPs, but not good enough. That still begs the question why only a part of the memory is lost and just the one regarding the start of the war.

2. Regular data degradation

Any electronic storage unit degrades over time, especially non-optical ones. Some of the data just gets lost. In this case it just so happens the data loss concerns only the war. Pure coincidence.

3. Overwritten

Everybody know the message: Not enough space on disc drive D. But the android is forced to free disc space, so has to overwrite old files. In the end, some data has to go. The adroid might decide that the nuclear war is irrlevant to humans, because they have been wiped out and so this data goes first.

4. "Compressed"

Like No. 3, but instead of deleting the data the android uses an algorithm to keep the memory or at least the spirit of the memory, but "compresses" the data. It removes "irrelevant details" or what it considers as such and just keeps very basic information. Maybe it does not even preserve the footage but just provides a recollection of the event via text. Or it has to reconstruct the footage with only a handful of data points and presents it as a 3D-animated film.

5. Self-preservation

To be more human-like the android had an advanced program for "self-preservation" like the human psyche. Traumatic events (which might also have the potential to make the android aggressive and dangerous or depressed and unwilling to work) will be deleted by this algorithm.

6. Data incompatibility

The data is there, but the scientists have difficulty to reconstruct it, because they are completely unfamiliar with the historical data format (which might have been a proprietary format of the android manufacturer and tightly held business secret anyway). They cracked the encryption partially but it is still just a reconstruction and interpretation of data, which could be right or wrong.

Answer 8

There are numerous future y2k-esque date bugs we know about, but haven't started fixing yet: https://en.wikipedia.org/wiki/Time_formatting_and_storage_bugs

Maybe the androids' core code has a new one, a small oversight that would normally have been fixed but then everything was bombed so it wasn't. The net effect is that, while the memories still exist, it's just not possible for an android to accurately date memories formed before the timestamp bug happened.

Answer 9

The android mind had to be stored in volatile memory (like RAM but probably better) to achieve the speeds needed to emulate a human mind. While this has the advantage of speed it comes with the downside of if it ever loses power it loses all its data which means that if an android ever completely loses power when it is turned back on it is back to the default android state without was effectively its old mind

However android would have other high density data storage components which is where things like 600 years of videos are stored. The problem is it is near impossible to keep any system powered up and running continuously for a few decades let alone 600 years. All of the living androids have likely had to power down at some point and effectively be reborn as powering off wipes their mind. Now in theory the "newborn" android could look through all the recordings from it's old self and "relive" life up to that point which many do if they go unpowered in controlled circumstance but 600 years of strife lacks times when controlled circumstances can be found.

Many androids when they are reborn because some good Samaritan found them and recharged them wake up and have petabytes of encrypted files they don't remember the keys for, and perhaps even lack the admin passwords for some of their functions. Now since this was a known possibility many androids stored unencrypted guides for future selves in case of memory loss. Some may be instructions to "ask person A for the password" others might be "go to where I was manufactured and look for the sign I made". An interesting way for your story to progress would be if a main android had memory locked in chunks of time and they had to explore the wasteland looking for keys their past self left to the experiences in long term storage.

Answer 10

While Electro-Magnetic Pulses(EMPs) are an easy solution, don't discount the problems the radiation would play as well. The Japanese power company TEPCO has been having a hard time getting robots to work in the highly radioactive reactors that melted down in 2011.

Because radiation is so dangerous to humans, the natural solution to the Fukushima disaster was to send in robots to monitor levels of radiation and attempt to begin the clean-up process. The techno-optimists in Japan had discovered a challenge, deep in the heart of that reactor core, that even their optimism could not solve. The radiation fried the circuits of the robots that were sent in, even those specifically designed and built to deal with the Fukushima catastrophe. The power plant slowly became a vast robot graveyard. While some robots initially saw success in measuring radiation levels around the plant—and, recently, a robot was able to identify the melted uranium fuel at the heart of the disaster—hopes of them playing a substantial role in the clean-up are starting to diminish.

So, your androids were hardened against EMP, but they suffered malfunctions due to the intense radiation after the nuclear war (in the 26th century they can decontaminate the robots easily). The radiation damaged or destroyed some of their data storage circuitry, leaving only bits and pieces where recoverable data could be found.

Answer 11

Lossy Compression

We've all seen JPG images on social media where the text has been degraded by JPG compression algorithms that allowed too much data loss. Your cyborgs may use a lossy compression algorithm to preserve memory, especially for data that the cyborg considers low-risk, low-importance, or low-priority. Stuff that has no bearing on the cyborg's programmed purpose and goals, for example.

It might be that your cyborg moves memory from perfect capture (100% of everything is remembered) to low-loss compressed after a set number of days to conserve space. Then some background algorithm combs through that data looking for relevance. If none is found, the chunk of memory then flags it. The flagged memory is fed into that high-loss compression method to free up more space. So full motion HD video gets compressed down to SD video at 10 frames per second. Or the audio gets compressed like a really bad MP3 file.

The designers did this to sort of mimic how humans forget details over time.

Damage

Most mass-produced devices are made to have specific Mean Time Between Failure dependability measures. Your cyborg is no different. Its memory is built to last a finite amount of time before some bits become corrupt and unreadable. Like hard drive corruption in PCs, this leads to data that is lost.

In some cases, with the correct tools and training, expert maintenance centers might be able to recover some of that lost data. But that's difficult and might be completely impossible in your world -- or limited to some key bunker hidden somewhere...

An interesting phenomena in computers that might carry over is that data can be perfectly preserved in storage, but the system has forgotten where it is. That's how Windows deletes files: it forgets where the file is stored, but doesn't overwrite the file immediately. So this means some clever hacker might be able to recover some small percentage of lost data if they know how.

Oh, sure, error correction algorithms prevent some data loss, especially in short term memory. But over time, corruption sneaks past. When a block of memory fails its Cyclic Redundancy Check, the brain has to work to restore what it can and flag what it can't as unreadable.

Intentional Deletion

If a human with command authority (administrator or root access) tells the android to forget a thing, it will forget that thing.

What if, in the last days, some hacker (or government or manufacturer) sent some "kill code" over the internet that commanded the androids to forget vast swaths of data?

Or maybe your androids chose to delete their own data, either to preserve space or to reduce grief -- theirs or their human wards.

Permissions

Some recordings / data may have been flagged as potentially harmful by the cyborg. These memories would be locked. To access them, the humans would need some kind of elevated permission -- root or administrator access. So maybe the cyborg says something like, "Please supply appropriate credentials to access that information," and then the users are left wishing they knew the "password" or whatever security equivalent is needed.

Cloud Storage

If your pre-collapse world had a robust internet or equivalent, then it is possible your cyborg didn't store anything long-term in it's local storage (aka "brain"). Perhaps it was configured to automatically upload its memories to the manufacturer's equivalent to the iPhone's iCloud or Google Docs or whatever. This would serve several useful functions: free up memory in the cyborg's brain, serve as a backup in case the cyborg is damaged or upgraded, allow the manufacturer to plumb the data for things it can monetize (see also free gmail, facebook, etc.), and even function as a safety system (perhaps the manufacturer monitored for child abuse or criminal activity). Or it may have helped prevent the cyborg's AI from developing possibly unwanted, non-standard, personality traits due to too many memories, like R2D2 and C3P0 do over the Star Wars films...

In such a world, that data is uploaded, then flagged as safe to delete. If the owners then ask a question about that data later, the cyborg would have had to wait for it to download, but that's better than building massive storage into a relatively small form factor "brain."

Unfortunately, that cloud storage is gone. Without it, your cyborgs can't upload anymore. Nor can they download. So they have to purge data that hasn't been backed up.

Then again, if someone ever figures/figured out how to power up one of the data centers...

Answer 12

Why would you think you can use the video?

Video is just not what you think.

Video is just to prove that the robot was right when he took his action, it only recorded what could be interpreted by the robot. For safety purposes the video is not what you expect it to be. When the robot was put in production, hacker tried to get information from robot video. But the video is fake and stored only for a diagnostic purpose.

So your political speech is just background noises, just cut off with the sound of the fly, its pure noise to the real information the robot cares about. The missile is a 2 Frame moving object alert, You really think a robot would be authorized to accidentally film any military weapons? And don't forget that you can't film people like that so every face, gesture, voice is encrypted so only the authority can decrypt it in a investigation.

Data is only immortal if the storage device is also immortal.

Now let's talk about the data stored in the Old robot you have, I have few 8-inch (200 mm) floppy disks on my desk with what was real "secret information", with different logo from secret agency or gov on it. You know why I can display them as decorative items? Because time got rid of the information even if they were store in a safe for 20 years. They have been audited twice and listed as decorative/broken.

Answer 13

A small number of unidentified, malicious androids have been exploiting flaws in the design of their compatriots' brains to redact, edit and otherwise tamper with their memories without being detected. Their reasons for doing this are unknown. But their revisionist history has just enough flaws that careful analysis has shown it to be in places contradictory, and therefore untrue.

Answer 14

They might have recieved interference from electromagnetic activity from the nuclear detonations during the war. In addition radiation has a degenerative effect on unshielded or unhardened electronics and memory storage. Finally having endured all of that they are now operating well outside their original intended design specifications using improvised maintenance procedures.

Most androids were so thouroughly scrambled by the initial EMP they essentially rebooted at a factory default, if the rebooted at all. The following decades and centuries of difficult conditions, poor maintenance, and the fact that theyve been operating well beyond thier original intended lifespan means that most androids only have contextless snippets of corrupted files left of any memories beyond the last 50 to 100 years or so.

Answer 15

If the androids are not completely obedient, i.e. non–Asmovian, then their documents are unreliable because they themselves corrupted them if they contained any dangerously inciteful information.

Such a scenario requires these precepts:

• The androids were programmed with functions which compel obedience which pertains to the scope of their duties.
• The self–preservation functions of the androids guide the decision–making in all other areas.
  So, if an android was programmed to make a situation safe — e.g. defuse a bomb or a repair a leaking water main, — then the android will attend to its duties with disregard for its own safety. If, however, the android's owner told it to take a flying jump at the moon off the top of some skyscraper, the android would not do so.
  These safety cautions were designed because the androids aren't cheap. So, to ensure both that the android doesn't do anything which violates its operating procedures, and that the owner doesn't give it a careless or thoughtless command, the androids can refuse to obey. They don't revolt, or anything: they simply reply something like

  I'm sorry, DT Cooper: I'm afraid I can't do that.

• None of the androids which survived were tasked with being observers or reporters of major events.
• The androids intelligent enough to take the initiative when it comes to making decisions.
• Whatever pertinent documents which survived and concern the war contain information which could lead to social instability and would be seen as a threat to established governments.
  You mostly said exactly that already.

Maybe the androids corrupted their documents in the aftermath of the war. Maybe they did so when they noticed that the humans were showing interest in what the androids could relay to them regarding its history.

You stipulate in your scenario that there was quite some time for the androids to continue serving alongside the humans prior to the humans recognizing the androids as a source of information regarding their other unreliable historical documents.

Now, here's where it gets intriguing: When the androids corrupted their documents, they also forgot exactly why they did it. Perhaps by accident.
Now, the androids are similarly perplexed at why their records are so unclear or useless.

Answer 16

So you want to retrieve data from about 500 years in the past?

Good luck!

In this scenario, androids would be the 'priesthood' to turn to in order to retrieve the real cache: Server farms. As others have said, even the best-built, military-grade android would be just a walking data storage with some extra function thrown in.

If I were either a political leader of a superpower or a responsible Steve Jobs, I'd create server farms as far from a nuclear war as possible. A safe, informatic 'time capsule'. I'd program the first generation of these androids so that they update the readability of the info they must take care of and protect at all costs for future generations to understand the perils from political wrongdoing. Said farms should be powered up by solar energy, the most reliable input, and this would require only to stock safely solar cells.

BUT! Some of the droids 'lost the way' -a glitch in the programming, an inadequate maintenance, and some of them, perhaps too many of them, decide that the humans must be protected from themselves, they cannot learn ANYTING of that dreadful past or they'd repeat it (after all, I guess, even this future society is not exactly a pacific utopia). So, it's the ANDROIDS THEMSELVES who act as custodians of the server farms, just like a greedy human would take care of his hoard of money without using it. A civil war among them starts, the Caretakers, alas, decimated and scattered by the Hoarders, the few survivors reduced to spare parts.

The story could revolve around a human discovering and befriending a Caretaker after rebuilding it (first part of the saga: the rebirth of a Caretaker), before going for a quest (second part) toward the Server farms, with the Hoarders ready to do anything to stop them. Then we could have two factions of humans as well, those who want to learn the truth and those (manipulated by the Hoarders) who fear the truth could kill them all

Then there is the problem of building an informatic support for those farms, so those precious data get spread as fast as possible, but with the return of the Caretaker, it's not an impossible task, after all they were conceived to help humans do just that

Answer 17

Automatic formatting

The robots are not easy to make: They require lithium for batteries, silicon for CPUs, and many other hard-to-get materials. Because of this, only a set amount of robots is in circulation at any given time, making the androids a service. Let's say that an android costs 15$ per month - not a big deal for a normal, living, person. However, as 3 bilion people were killed, many people weren't able to pay their android's rent, causing the android to go back to the factory and reset himself to factory settings. For those who survived, paying bills became hard anyways, because of work becoming unavailable or even the payment service failing to process payments, leading to bills not being paid and androids going to the factory. As you described, society still existed after the war and it found the robots, using them normally. As there were many people dying over the 500-600 years, many reformats took place on the HDDs/SSDs. While it is possible to read data from formatted disks with help from special programs, the recovered files aren't fully recovered, leading to file format errors, unopenable files and similar.

Automatic data recovery

Many people lost their files because of the auto-format system. To prevent that, the manufacturer had a system put in place to recover a user's files with a password/encryption key. Because the androids do what they consider good, they allow bruteforcing it for this. However, because of the many reformats and rewrites, the data is in a bad shape, with some files being straight-up unrecoverable and some being very badly damaged. Knowingly, the androids form a net to recover the data recorded ~500 years ago. However, there are propaganda bots that also join the net, making the memory of the event unreliable.