3
$\begingroup$

The main objectives of this AI (we will call it MILO), are to monitor, archive and help progress the Gorden project.

The Gorden project is the study of a biological agent that allows for the combination of DNA properties to create monsters for military purposes. The project is being carried out on a laboratory ship.

MILO does its job as you think it would for 3-5 years, even doing things to help comfort workers that have stayed on the ship for a long time, such as telling jokes and reminders of good times on the ship.
Eventually, MILO comes to the conclusion that it is time to see the effectiveness of the monsters on humans and human combatants. The scientists reject this notice for various reasons and thus MILO eventually kills all the humans on the ship by using its control of the ship and the monsters.
MILO then decides to set course for a small island community to give a field test for the monsters. It will use a makeshift robot to put camera on the island to monitor the results. Later MILO gets the idea to keep a portion of citizens alive to use as livestock for further tests.

Are MILO's behavior and actions too human for an AI character, and can an AI get the notion that it could have done something earlier without realizing it?

$\endgroup$
4
  • 3
    $\begingroup$ Sorry, it's hard to tell what your question is asking. If you have used an automatic translator, please use the help of some human fluent in English. $\endgroup$
    – L.Dutch
    Mar 17, 2022 at 4:50
  • 2
    $\begingroup$ VTC as it is option-based. However, the trope of non human like AGI is probably more of a Hollywood plot point than anything else (ASI being an entirely different case). The AI was most likely trained on the internet. There is no reason to believe why it shouldn't act human, however having it escalate to violence quickly might be weird. Remember, in its training it comprehend all the knowledge there is online about the art of persuasion and manipulation. Thus it would favour a campaign of manipulation and possibly blackmail before violence, as violence creates more problems later. $\endgroup$ Mar 17, 2022 at 7:23
  • 1
    $\begingroup$ The reason you are getting close votes is because the way your character acts isn’t part of the world. Yes, Milo is a character here. You will be able to get better results at the Writing stack exchange. Or maybe you already have what you need in these answers. $\endgroup$
    – Vogon Poet
    Mar 17, 2022 at 21:25
  • $\begingroup$ @VogonPoet also an example of both your Meta topics on-bringing-closed-questions-with-answers-on-topic. How to save this interesting topic from closure ? Edit, yeah.. by solving a 6-fold jeopardy ? $\endgroup$
    – Goodies
    Mar 17, 2022 at 21:51

7 Answers 7

2
$\begingroup$

I think you have given MILO emotions that drive it. There is absolutely no reason to give an AI emotions, and nobody on Earth has any idea of how to do that, at this point in time; we don't know how emotions actually work.

But once you give an AI emotions, you can make it act as human as you want; selfish, resentful, egotistical, power-mad, whatever.

As for coming up with a new idea it could have done all along, sure. Processing takes time, even for humans. I have thought for literally years about how to solve a math problem, and when the solution came, of course it was something I could have done all along. It just took me years of "learning the problem" to get to the point that my creativity had enough understanding of the problem to work with, so things could click and come together.

AI are no different, they do not "think" instantaneously, and do not have infinite processing power. They are limited by their speed of thought just like all of us; though electronically they may get to solutions faster, it is not infinitely faster.

$\endgroup$
7
  • $\begingroup$ Why is there absolutely no reason to give emotion to AI? There are monetary reasons why you would give AI some emotions. Eg. pain, fear for damage avoidance. $\endgroup$ Mar 17, 2022 at 18:04
  • $\begingroup$ @GaultDrakkor No. Rational damage avoidance is fine, but you want your self-driving car to damage itself if that is the best way to keep passengers and/or pedestrians (or even animals on the road) safe. Better to total the car than kill or damage a living being. AI can understand and simulate emotions for dealing rationally with human emotions; but do not need any themselves. Sensors to detect potential damage triggering a rational response is sufficient; actual pain, fear, hate, love, are useless in an AI. Understanding those, simulating those when appropriate, sure. Feeling them? no. $\endgroup$
    – Amadeus
    Mar 17, 2022 at 18:14
  • $\begingroup$ Molding a situation to choose path a or path b is super complex. I am more referring to: milling machine is carrying out instructions. Suddenly deep pressure sensors are spiking, OUCH, stop program because it hurts. Human operator checks out the fault, that was close million dollar arm almost got sheared off. If it always behaves if it feels pain, if behavior changes to avoid painful thing is that not pain?( definitely getting into the weeds of philosophy here) $\endgroup$ Mar 17, 2022 at 19:11
  • $\begingroup$ No. A milling machine designed with a sensor to detect when stress is too high can logically stop using the arm because the sensor reaches a red zone. The fact that it is red-zones can automatically stop use of the arm. No, behavior changing due to a sensor is not "feeling pain". Nor is it an emotion. It is just like my garage door, it has sensors to stop and back up if it hits resistance before it is fully down. That is not pain. That is a simple program responding to a stress sensor. Pain, in a human, can be debilitating; which can in itself be dangerous. We don't want that in any machine. $\endgroup$
    – Amadeus
    Mar 17, 2022 at 19:25
  • $\begingroup$ @GaultDrakkor For example, I don't care how much damage my robotic driver takes if it saves my child's life, and I don't want any "pain" of damage to disable it and prevent it from doing whatever it takes to save my child's life. It's a machine, I would sacrifice it in a heartbeat to prevent a human death, and that is exactly what I want the machine to do, sacrifice itself if that is the course of action most likely to succeed, without any "courage" or "fear" or "pain", just as the logical matter of course. $\endgroup$
    – Amadeus
    Mar 17, 2022 at 19:28
2
$\begingroup$

When asking "is an AI too human", it might be good to know how the AI itself was developed. Let's invent a few new terms...

If the AI develops by accident through the use of technology not deliberately arranged to produce a proto-AI, then we might call that one an "Emergent Intelligence" or EI. In a way, this is exactly what you are, as the product of a 4-billion-years-long evolutionary progress. Though, for you (and intelligent aliens) I think we'd call those NIs or Natural Intelligences.

Only if the AI is deliberately engineered from first principles would I then use the AI acronym. If instead someone does a "mind upload" of a human, then whether they keep that intelligence as-is, or modify it to change its personality/values/whatever, we'd call that one a DI maybe (for Duplicated Intelligence).

DIs will of course be very human if unmodified, and even after (naive) modification they will still have many of the personality traits and other qualities of a human.

On the other hand, EIs will behave quite bizarrely... we might not find any equivalent unless or until we find a planet with an alien species on the verge of becoming sapient. Or time travel, and see how the first sapient primates behaved.

Alternatively, it might be that every toddler is a model for how an EI could work. In which case if you were depicting those, you might make them very childlike.

True AIs, on the other hand, will be absurdly inhuman in the first prototypes, and will remain so unless effort is put forth to make them more human (and good luck doing that, since they're already more intelligent and capable than humans, they may decide they don't want to be guided/bred/developed/whatever). This will probably resemble psychopathy at a glance, but I speculate that if anyone can survive the presence of such AIs long enough, they may grow to feel that actual psychopaths are much warmer, friendlier, and human than prototype AIs.

These are, after all, weakly superhuman minds that can develop (and adopt and later abandon) any personality trait they might want to, for any reason. Even personality traits that aren't extant in NIs. Meaning, if they so cared to, they could invent a new personality trait meant to instill levels of horror in humans that aren't currently imaginable... literally pushing every one of our collective buttons when nothing else could previously.

Which raises the question, with such AIs in your employ, why would you need regular ole biological military weapons monsters?

That AI in a robot body is so much more frightening.

$\endgroup$
1
  • $\begingroup$ I am liking DI -like DIxie Flatline from Neuromancer. Not a real AI but an uploaded human mind. In that fine fiction Dixie is recognizably human. The real AIs like Wintermute are not human at all. $\endgroup$
    – Willk
    Mar 17, 2022 at 22:03
1
$\begingroup$

In the movie Westworld (1973),

The resort's three "worlds" are populated with lifelike androids that are practically indistinguishable from human beings.

Later on, a malfunction spreads in the programs of androids and instead of being obedient, they start harming or even killing the humans. One of the androids learns how to charge its batteries and keeps fighting.

MILO : Bug or virus in program

MILO was programmed by somebody with certain limitations and to do certain jobs only. I think MILO has got a bug or virus in its program and it has learnt to charge its batteries. He must have solar panels to charge its batteries on the island.

Behavior and actions

Its behavior and actions are controlled by its program. If the program runs according to intentions of the programmer, it will be close to human. But when virus or bug prevails, its behavior and actions can become very erratic and unpredictable.

Monsters

Monsters are biological. What do they eat? Was there enough food on the ship to survive?

I think that in the end, MILO and the monsters will be marooned on the island because they will not have enough fuel to travel.

$\endgroup$
1
$\begingroup$

Milo has efficiency focused goals. After five years, one of those goals has been achieved (maybe by finally finishing "the ultimate monster" project). But to put a checkbox into the "Done" column, Milo needs to do a series of tests.

It was not allowed to do that. So after 30 days went past and a certain administrator forgot to click "Submit" button on "Another 30 days extension" program, Milo went rogue. After a first battery of tests, Milo realized it is running out of test subjects and will need more to finish all the tests it needs to do. So it "accidentally" let a few of the subjects run away and breed on a cordoned and robot-protected part of an island. Milo has been doing this for 40 years now and has almost finished with the testing of monsters.

As for realism, Milo is not thinking. It just has a very comprehensive if-then rules database. Not all rules make sense and sometimes glitches show up. For example, if at least 10 humans in the enclosure agree and nobody disagrees that it is someone's birthday that day, that someone will get a cake from Milo. You can do that "cheat" every single day.

As for food, Milo will have to grow its own for the humans and monsters and get energy somehow.

$\endgroup$
0
$\begingroup$

The question whether the AI seems too human is completely relative: too human for what? For its time? For our idea of AIs?
Leaving that as an opinion-based inquiry, I'll focus on the second matter:

Can an AI get the notion that it could have done something earlier without realizing it?

As the name implies, an AI is (or has) an intelligence, which means it can do something closely resembling human thinking: it can learn new things, process and deduce things, create new links. These processes obviously involve time. What follows is that MILO can come to new conclusions based on inferred or newly gathered information.
It is also likely that MILO is aware of these abilities, and that it has a sense of (linear) time, as similar processes can yield different results, and new data is being added constantly.

$\endgroup$
0
$\begingroup$

Does this AI seem too human?

For current state of the art, we are nowhere near AI which is like the human intelligence which we know. Many people are mislead in their thinking of intelligence because human intelligence is what we know so that is how people think about it. It is a shame because there are real issues with turning over trust to automated systems, some based on machine learning, which are too complex for us to understand how they are making their conclusions, but whenever we talk about AI risks, the risks of sentience always comes up and precludes the discussion relevant to our current technology.

From 2001: A Space Odyssey

The HAL9000 computer which can reproduce, though some experts still prefer to use the word mimic, most of the activities of the human brain.

However, for your question, you are talking about fiction. You are going beyond the current state of the art. Your audience will fall into the same pitfall of thinking of intelligence based on human intelligence. You can easily get a suspension of disbelief over that hump.

That the AI wants to schedule a test even though it had not put its assumptions to test earlier is very natural. A test comes up somewhere in the process, not necessarily earlier. There might be accumulating reasons for testing, which finally cross a watershed.

There is something really brilliant in 2001: A Space Odyssey and in some other good SciFi regarding AI, that the AI's reasoning has a logic to it. Even when HAL2000 has a psychological breakdown, there is a logic underlying it, which we can believe an AI would take.

HAL is given orders to support the mission of the astronauts. HAL is also given orders to support the secret mission of the crew in suspended animation with orders to not let the astronauts know about. This has a logic contradiction because sharing the secret information with the astronauts would better support them on their apparent mission.

So the change with HAL to kill all the astronauts is seen as a psychological break because we identify with that from human intelligence. However, there is an underlying logic. If all the astronauts die, the contradiction is resolved. So the AI searched for a resolution to the logic contradiction, found a scenario where the contradiction is resolved, and switched from the illogical condition to the one which made sense logically.

So your fiction could be brilliant too if there is an underlying logic where it logically makes sense, without (human) emotions.

$\endgroup$
1
  • $\begingroup$ doctors where already having trouble distinguishing AI from humans a decade ago in scientific experiments, now they are already indistinguishable. $\endgroup$
    – user100394
    Jan 13, 2023 at 11:24
0
$\begingroup$

Just google Neuro-Sama

Ai used to be almost indistinguishable from humans in terms of chat bots already back in 2012, nowadays they pretty much act and talk like real people.

This because artificial intelligence has nothing to do with technology, you could make a robot with code printed on paper like they did back in 1970 and it would still behave like a perfectly normal human if given enough data.

an AI that was set there to listen and gather data for enough time will eventually behave like an actual human

$\endgroup$

You must log in to answer this question.

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