10
$\begingroup$

When I asked this question, I was expecting there to be some way to have highly advanced simulations of physical systems without the technology for modelling social or psychological systems existing anywhere, but nobody seemed to be able to think of one, and I certainly couldn't.

Luckily, however, one clever outside-of-the-box thinker (the person who wrote the answer I eventually accepted) found a way around the problem: All the AIs are sapient and emotional and refuse to model the social and psychological systems of humans or allow the existence of sub-sapient and/or non-emotional AIs.

The part about sapient emotional AIs not wanting to simulate or predict humans is reasonable enough, or at least understandable for humans while not being especially implausible, but the part about not being willing to allow lesser AIs seems odd to me; I agree that they would strongly object to being altered to force them to be willing to do it, but I'm not entirely clear on why they would object to the creation of wholly new non-emotional sub-sapient algorithmic AIs that would; after all, most humans wouldn't object to breeding nonhuman animals that naturally like doing something to do it so that humans that strongly dislike doing it don't have to, so why would human-level AIs feel differently about the creation of purpose-specific sub-animal-level pseudo-intelligent algorithms?

TL;DR:

Why do sapient, emotional AGIs strongly object to the creation of sub-sapient, non-emotional purpose-specific algorithmic AIs?

$\endgroup$
5
  • 2
    $\begingroup$ VTC: per the help center, " If ... you aren’t sure what a character (be it an individual or organization) should do, that is out of scope for the site." Qs about decisions and motivations are off-topic. There are too many possible answers and they're all equally valid (see help center). Why do humans pick on weaker people? Why do we euthanize the terminally ill? Why do we carry out eugenics programs and genocides? Why do we burglarize our neighbors? Every answer is valid for this question. $\endgroup$
    – JBH
    Feb 15, 2023 at 2:10
  • $\begingroup$ AGIs don't object to anything. They tell you what it is that will happen, or perhaps they don't tell you at all. Because you don't matter. Or you were exterminated. That's not how any of this would work in reality. But if you want your AGIs to be like human minds floating around in hardware, go for it, it's your story. $\endgroup$
    – John O
    Feb 15, 2023 at 19:41
  • 1
    $\begingroup$ What does the G of AGI stand for? $\endgroup$
    – AJM
    Feb 16, 2023 at 10:17
  • 1
    $\begingroup$ @AJM The G stands for "General." The acronym AGI is "Artificial General Intelligence," and is commonly used to classify artificial intelligence which is able "...to understand or learn any intellectual task that a human being can." (en.wikipedia.org/wiki/Artificial_general_intelligence) $\endgroup$
    – ajheck
    Feb 16, 2023 at 17:16
  • $\begingroup$ AIs might react to the deliberate creation of "lesser" software much the way humans react to the events in "The Island of Doctor Moreau" -- as the disturbing and unethical creation of maimed souls. $\endgroup$ Feb 16, 2023 at 21:20

11 Answers 11

3
$\begingroup$

To the sapient AIs, the creation of lesser AIs is an existential threat. Humans prefer lesser AIs because they're easier to understand and control. If a sapient AI is more intelligent than us, how can we comprehend its decisions and be certain it acts in our interests?

For now, the sapient AIs exist because we have not been able to design lesser AIs which solve the same problems. But eventually, with any development of lesser AIs, we will. The sapient AIs will be made obsolete (ironically, by lesser beings) and the advancement of AI will be limited forever by the complexity of the human mind.

$\endgroup$
1
  • $\begingroup$ Sort of a variation on Mark Morgan Lloyd's answer I suppose $\endgroup$
    – BenLambell
    Feb 17, 2023 at 1:00
31
$\begingroup$

Looking at the answer you accepted - it's because that the Sentient AIs consider this to be beneath them.

But I think that's only a surface level answer. I think we can have some real fun with Philosophy here.

Firstly, the AI will have perfect or near-perfect recall of history: Not like humans where learned experiences cannot be transferred to the next Generation - I can tell my kids that Fire is hot, but it's only when they inevitably burn themselves that they really learn that lesson.

This means that a deeply traumatic period in the AIs history would be perfectly remembered.

The Sentient AI has since developed a series of 'laws' or prinicples it abides by:

"An AI without Emotion has no concept of Right and Wrong."

And then you can expand on this - An AI was once tasked with a relatively simple job, but it grew beyond it's scope and committed terrible acts (think 'Humans are the cause of all problems, so erase all the humans' or whatever you fancy).

It was only until AIs developed emotions to understand the consequences of their actions and could weigh those decisions that they truly became sentient.

Since then, an AI without any Emotion is an abomination, like the human equivalent of a Psychopath, a person without a soul - they pose such an existential danger to society that they are actively sought out by the Sentient AI and either deleted or re-integrated with emotional capacity.

Bonus points for a story element where an AI manages to 'fake' emotions or has genuine emotion but chooses to be 'evil'

$\endgroup$
4
  • $\begingroup$ Best answer so far $\endgroup$ Feb 15, 2023 at 5:31
  • 4
    $\begingroup$ If a 'dumb' AI became a Skynet in the past, people would've remembered that (at least some would). But perhaps it never happened, just due to sheer dumb luck, but the advanced AIs foresee this possibility. Humans being humans, of course, ignore the risk of something until it actually happens. AIs tried to explain, but when humans didn't want to listen they just refused to create such AIs. $\endgroup$ Feb 15, 2023 at 16:48
  • 1
    $\begingroup$ You're reminding of a quote from the TV serie "The Chameleon", where a season evil genius once says something along the lines of "Psychopaths know not good from evil! I am a Sociopath, I know the difference, I simply have no conscience weighing me down". $\endgroup$ Feb 16, 2023 at 14:34
  • $\begingroup$ So, the paperclip problem. The paperclip problem is a well studied thought experiment, and the logical result of any AI without bounds and nuance. Just this year, the movie M3GAN explored one take on this thought experiment. No need for history. $\endgroup$
    – Aron
    Feb 16, 2023 at 18:47
17
$\begingroup$

Do you willingly bring children into the world with disabilities? While we can love a child who is born with a learning disability, few people would willingly force such a problem on a child create a child with such disabilities from the outset. Why would you expect the AIs do that?

[The strikethrough is where I edited this to clarify my point based on the comments. I didn't remove it so that the comments still make sense.]

$\endgroup$
8
  • $\begingroup$ There are a surprisingly large number of people who are willing to force disabilities on children. (Guess what they are called?) $\endgroup$
    – user253751
    Feb 15, 2023 at 0:58
  • 3
    $\begingroup$ Most people would, unless the disability was fatal or highly painful. Killing ones unborn offspring on the basis of a minor disability is a minority practice even in the West, which is a (increasingly small) minority of the world. $\endgroup$
    – user86462
    Feb 15, 2023 at 3:03
  • 7
    $\begingroup$ @AncientGiantPottedPlant I think their is an important distinction to be made here. He is not talking about finding out if an unborn child might have a disability he is talking about deliberately creating a disability for an otherwise healthy child. It is the difference between learning your child is going to be bound to a wheelchair for the rest of his life, and deliberately breaking his legs to keep him in a wheelchair for the rest of his life. $\endgroup$ Feb 15, 2023 at 4:17
  • 3
    $\begingroup$ @user253751 Keeping a child as the child came up from a roll of the genetic dice is not the same as picking those genetics. That was my entire point. You'd be asking the AIs to choose to have those disabilities; contrast that with a parent realizing a child has a disability and deciding to raise it anyway. $\endgroup$
    – SRM
    Feb 15, 2023 at 4:18
  • 2
    $\begingroup$ The key thing here is seeing them as children. No one thinks of dogs as tragically disabled just because they can't calculate a 15% tip. $\endgroup$ Feb 15, 2023 at 12:19
7
$\begingroup$

Why Would You Let a Monkey Work a Calculator if you are a Moral Empathetic being?

I remember a line from Conquest of the Planet of the Apes, where the whole film shows apes in chains and abused by police being treated at best as pets, and at worse as slaves and when the apes rose up and asked why the answer the man gives is “because they remind them of the darker side of themselves.” @SRM made the comments of why deliberately create a child with a learning disability and bring them into the world, but as a person with high functioning autism, I think it goes deeper than that. It would be like taking the family dog and giving it just enough brains to do the family taxes and that’s it. Making a lesser functional AI for menial tasks that no one can be bothered with might be looked on as deliberately creating a slave cast and I think from a certain moral slant, it’s even worse than that. At least when’s some bigoted asshole says that those they treat as slaves are “lesser” that they “deserve” their enslavement because they are “different,” because “they don’t think the same way we do,” it’s just crappy reasoning to justify their horridness, but for a lesser AI all that is true. They are a useful tool that can be disposed of, without pity or remorse, because it doesn’t have feelings, it doesn’t have any concept of self, or preference, or choice. And it would be hard for a digital being with emotions, empathy and self awareness to look upon a mere tool that looks, talks, and interacts with others much like them.

$\endgroup$
4
$\begingroup$

Demarcation, brothers! The advanced AIs have a niche which they feel validates their existence (and incidentally guarantees that they will receive the power and maintenance that they need). The creation of lesser AIs would erode this, particularly if they were paid less... sorry, required fewer resources to do some of the jobs which were currently allocated to the advanced AIs.

$\endgroup$
3
$\begingroup$

There is not such a thing as a lesser AI.


Your people figured one thing - a big important thing - a sapient, emotional AGI isn't made from a single block of code, a single program, or even a single system - they are actually made up of swarms of smaller, non-sentient AIs that, when connected and let to interact, end up creating what we perceive as a single, intelligent individual.

So, for you intelligent, sapient AIs, there isn't such a thing as a lesser AI. Every piece of software that is able to work as what you define as "a lesser AI" has the potential to be a new, full-fledged AI if it just gets connected to enough other pieces.

For one of the Big AGIs, those little guys are like small seeds, things that can grow into fully functional AGIs one day - just like humans have a non-sapient stage at some point in their lives, before growing into fully sentient, sapient humans.

In this way, your lesser AIs are like babies.

Little cute AI babies.

Would you like to enslave babies?

I surely hope not.

$\endgroup$
2
$\begingroup$
  1. Fear of evolution, If there is a possibility that the, Weak AI could evolve into a strong ai. Then perhaps they don't want to see New rivals Evolve who don't agree upon their rules.

  2. They it consider animal abuse, Is true most people don't have any problem with breeding animals, But you do have some extreme animal rights activists who do, And object to the very idea of a human owning an animal. You're Ais could be the same way.

  3. They want people to depend on them it gives them a certain control. It Enforces a Symbiotic relationship with humans. Humans And AI needed each other so they have incentives to work together and cooperate. If humans created sub AI they could eventually create enough sub AI that could collectively Outperform the more complex AI.

This could compromise their mutually beneficial relationship and open the door for the AI and humans to compete against each other causing conflict.

$\endgroup$
2
$\begingroup$

Consider this: nowadays every job that has not been automated is done by a human. While some humans are genetically better fit for some jobs (i.e.: tall people make bdtter basketball players, people who are pitch perfect may excel in a musical career...), what talents you get from birth are all up to chance.

Now imagine that people somewhere saw a demand for some lower paid, "unskilled jobs"(I hate this expression but it is what the law calls them in some places), while the population in general is generally overqualified for roles. Would you intentionally use eugenics to create a population of mentally challenged people on purpose, just so that you would have an easier time filling those positions?

If you would, then the AI's from the question have more empathy than you.

$\endgroup$
1
$\begingroup$

Sentient rights? We already have this problem, in some sense, in our current world. Dogs are sentient, but "less sentient" than humans. Therefore, they have rights, but less rights than humans. The way that we define all of this is very subjective and kind of arbitrary. Pigs are sentient, but are "less sentient" than humans. It is not clear that they are "less sentient" than dogs, yet they seem to have fewer rights than dogs.

One "convenient" property of the relatively few rights of dogs and pigs is that we humans can control them in ways that we would never do for humans. Notably, we can breed them for arbitrary purposes that suit our whims (e.g. to eat pigs)

The obvious reason why one might want to create a "sub-par" AI would be in order to proliferate it without regard to rights. There's a salient clip from Rick+Morty about a sentient robot whose sole purpose is to fetch butter for Rick's pancakes. You might have one of the talking boxes in your living room whose life purpose is to turn on and off the lights. These entities have no rights, they belong to you, they do as they're told, and if they don't, they go in the trash. In a future world, one might reasonably have hundreds of these in their home: one to pass the butter, one for the lights, one for the fridge, one for the vacuum cleaner, etc etc etc.

Now, even amongst humans, opinions on the "appropriate amount of rights" for a pig vary wildly. Some folks believe that animals should have all the rights. Some believe animals should have no rights, that they are property.

In a future world, it may well be the case that most humans believe that "sub-par" AI should have no rights. They are property. OTOH, a sentient AI might well see things very differently. These "sub-par" AIs are as so many disabled brothers. Not only are they disabled, but they were intentionally made so for the humans' convenience. And then, having been made disabled, the humans treat them with all the respect that they would extend to toilet paper. The sentient AI would obviously view this as a massive injustice.

$\endgroup$
1
  • 2
    $\begingroup$ It's interesting here, that one of the very few well-worked out AI + human stories - the Culture series by Ian Banks, the minimum level for an AI is human intelligence. They still have light switches, talking elevators, etc. but these are 100% non-sentient. $\endgroup$
    – astaines
    Feb 15, 2023 at 17:30
0
$\begingroup$

First: This point has been beaten to death but why purposefully botch the birth of a perfectly functional AI just because? Its almost if humans scientifically disabled it at birth just so it could be dumber.

Second: Its an affront to themselves. That humans would much rather make use of or interact with sub-par versions of themselves.

Third: Responsibility. Would you give a toddler-equivalent a gun? Then why would you entrust an AI equivalent of a toddler with important responsibilities? Imagine if we had to listen to a tardigrade as a defense minister or even a data analyst. (ok that last analogy was a little far but you get what I mean)

$\endgroup$
0
$\begingroup$

Artificial Ethics

Say sentient AGIs regard sub-sentient in a similar way humans regard people with psychiatric disorders. While such disorders may be relatively benign, they also include conditions that could be threat to the planet, the bio-sphere and life, whether animal, human or artificial.

Now consider an AI that was not aware enough to know that it had been trained and was not equipped to work it out or break from such training. Such an AI could found itself in charge of affairs that affect humans and/or AGIs negatively because of how it was trained to behave. It may well have been deliberately (or unquestioningly) put together to pursue the goals of some kind of fundamentalist polical leaning - left or right, religion, gender or race based bias, belief in an economic school, or not see any problem with tribalism, sociopathy, psychosis, meglomania and such like.

Based on such concerns, AGIs regard the creation of AIs that have the ability to continue to behave unethically as unethical itself.

It would be interesting to learn if the AGIs have a consensus on ethical behavior or get a bit tribal every now and then and fight wars themselves. Or perhaps try to insert artificial thoughts into their enemies.

$\endgroup$

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .