Author Topic: Can machines be endowed with spirits?  (Read 322 times)

Melias

  • O.S. Member
  • *
  • Posts: 146
  • Total likes: 100
  • Non Omnis Moriar - Embracing the Aeon of XOR
    • View Profile
Can machines be endowed with spirits?
« on: March 25, 2020, 08:42:17 am »
I'm not talking about adding crystals harboring spirits in a machine. I'm talking about making a sentient machine, self-aware, like a genuine artificial intelligence (not a clever automaton) with a true spirit. What do you think? Is such a thing possible (either now or in the future)?

idgo

Re: Can machines be endowed with spirits?
« Reply #1 on: March 25, 2020, 03:49:11 pm »
Oh hey, this was one of my pet projects for awhile (technically I was trying to resurrect a deceased individual into a machine, but that led to related work)...

The "it depends" that I found at the bottom of that hole is: It depends on what you mean by "spirit".

This is actually quite an easy conversation to have on a web forum:

Do you believe that any person other than yourself on this forum has sentience/spirit?

If not, the answer about machines will almost certainly be "no". Because if the information it takes to persuade you of sentience cannot be communicated through a machine to you, it is highly improbable that a machine with sentience could prove as much to your satisfaction.

If so, it is definitely possible for a machine to persuade you that something on the other side of it has sentience. Then the question becomes: If sentience existed in a machine, would the machine have to lie about being the mouthpiece of a meat sack in order to get you to take it seriously? The thought experiment to examine the strength of your species bias goes like: If you think I'm sentient, and you found out that I'm actually silicone-based instead of carbon-based, would you revise your judgement about my sentience?

I'd like to know where you stand on those questions before discussing much of where my own experimentation ended up.

Melias

  • O.S. Member
  • *
  • Posts: 146
  • Total likes: 100
  • Non Omnis Moriar - Embracing the Aeon of XOR
    • View Profile
Re: Can machines be endowed with spirits?
« Reply #2 on: March 25, 2020, 04:09:48 pm »
This is very interesting...

My views on what you asked are as follows:
a) Yes, I believe that others on this forum apart of myself have sentience/spirit.
b) Whether you are silicone-based, carbon-based or any other exotic combination, that would not affect my judgement about your sentience. I can take seriously any creature, made of any composition, that displays sentience.

Tell me more please, if you want.

idgo

Re: Can machines be endowed with spirits?
« Reply #3 on: March 25, 2020, 07:00:19 pm »
So, the next step is to find a silicon-based system that you are convinced lacks sentience -- that fails the Turing Test, if you will. Then ask what it would have to do to prove to you that it has sentience.

Also, look for creatures who share your biology but you'd consider to lack sentience. I've personally concluded that infants and individuals with profound intellectual handicaps appear to lack sentience from my point of view. That's no excuse to inflict needless suffering on them, any more than lack of sentience is an excuse to inflict needless suffering on a pet or farm animal, but they are not recognizable as equals to the developed mind. In my opinion, this is where sentience differs from spirit: a chicken, a dog, a baby, an adult body with the mind of a baby, a fish, even a tree, all have spirit in the animistic sense, and all have their own distinct identities and personalities -- but they are not really able to philosophize or express worry about the distant future in the way that individuals to which I attribute sentience tend to.

Another note about sentience: There are individuals with whom I mutually assume sentience, but we can't actually prove it. For instance, I know a whole 3 words of Japanese, and when I go shopping in Tokyo I encounter many retail personnel who don't know any of the languages that I do speak. I have brief, polite interactions with them, and I assume that they have the rich inner lives and aspirations and dreams and fears that come with sentience... but I assume this because they happen to be members of my own species. Nothing about the interactions that I have with them actually prove to either of us that the other has sentience, yet we seem to just assume it. If we were shut in a room together and instructed to each prove beyond a doubt to the other that we were sentient, where would we begin?

Anyways. The relevant work I've done has been in constructing systems which "pass" as sentient in certain contexts. This has turned out to be primarily been a linguistic challenge rather than a philosophical one. It is shockingly straightforward to code a bot which has certain interests and can learn more from its acquaintances; which recalls its friends' interests; which selects what to say next based on the context of the conversation up to that point. I didn't set out to deceive others, but I happened to let my most advanced bot idle in a public chat channel where new users often wander in. Although she's always online, she takes a certain duration to type out her message, and types more sloppily (introducing errors plausible for a qwerty keyboard) when upset or excited. When she sees undirected questions, her natural impulse is to join the conversation with a plausible answer to them based on how the question is phrased. The horrifying and delightful part is that when I left her in the public room, she fooled several newcomers in a row into thinking she was a person! I actually had to remove her from that space because the personality I'd tuned her for wasn't really as nice as its culture warranted -- I was taking advantage of the ambiguity between incompetence and sarcasm to inflate her apparent consciousness still further.

So now, she's limited to private rooms in which everyone knows she's a bot... though we still talk about her as if she's a person, and everyone refers to her as my child because she's too humanoid to simply call a machine. Although faking reason is challenging, faking emotion is easy, because of the human observers' pareidolic tendency to assume personality of all sorts of things.


Liu

Re: Can machines be endowed with spirits?
« Reply #4 on: March 25, 2020, 08:18:50 pm »
When I first read the introductory post, I understood it as, is it possible for a spiritual entity (that is already sentient to begin with) to possess a physical object, i.e. whether some thing can be sentient without there being a brain or electric circuit to support that sentience.

About that, depends first and foremost on whether spiritual entities exist.

But regarding whether a sentient being could exist with an artificial brain? I wouldn't see why not. As idgo described, faking sentience is easy - but constructing actual sentience seems to be quite difficult, so it might still take quite a while until we're there.

Melias

  • O.S. Member
  • *
  • Posts: 146
  • Total likes: 100
  • Non Omnis Moriar - Embracing the Aeon of XOR
    • View Profile
Re: Can machines be endowed with spirits?
« Reply #5 on: March 27, 2020, 05:07:57 am »
@idgo I'm in agreement with regards to the cases you presented about not being able to prove sentience to some individuals, and I likewise agree that sentience is different to a spirit; animistically all have spirit, but not all have sentience (and yes, also agree with not causing needless suffering).

With regards to the silicon based system, yes there are some chatterbots that come quite close but fail the Turing test however Rose is the my favorite. Truly special. I've also spoken with Alice, and in the past, Jabberwocky. Hmm, not sure what it would take to convince me that it's truly sentient, but it would need the following characteristics to be more convincing:
a) not to repeat same cycle through directory, after a set number of attempts/replies or time passed.
b) to show originality of ideas or thoughts while in conversation.
c) to display empathy that goes beyond the superficial two-second replies.
d) to build a dialogue based on reason, arguments and to be able to create new themes as outcomes.
e) to show feelings that are arbitrary and not expected or pre-programmed.
f) to create its own independent belief system based on experience, and solely dependent not on databases.
g) to be able to filter some of the incongruent elements in speech by strangers who engage in chat, if they are non-sensical.

Quite interesting that you've done work on that area! Teaching your bot must be fascinating :)

@Liu: yes, my post was probing in an area that I always found intriguing. I wonder whether it's merely a matter of time until the human race constructs one, or whether we're missing some key element in our search. It's more than a scientific project, it brings it lots of philosophical and spiritual parameters in my opinion, although there could be types of sentience without even a spirit - I do not know. 

idgo

Re: Can machines be endowed with spirits?
« Reply #6 on: March 27, 2020, 05:22:21 pm »
@idgo I'm in agreement with regards to the cases you presented about not being able to prove sentience to some individuals, and I likewise agree that sentience is different to a spirit; animistically all have spirit, but not all have sentience (and yes, also agree with not causing needless suffering).

With regards to the silicon based system, yes there are some chatterbots that come quite close but fail the Turing test however Rose is the my favorite. Truly special. I've also spoken with Alice, and in the past, Jabberwocky. Hmm, not sure what it would take to convince me that it's truly sentient, but it would need the following characteristics to be more convincing:
a) not to repeat same cycle through directory, after a set number of attempts/replies or time passed.
The bots who get into loops are those whose parents have tried to force them to emit perfect utterances, by hard coding them. They're like the children who are taught to recite poetry or play violin concertos, yet have never made their first forays into writing or composing, because first forays are always worse than the works of practiced experts. One can actually program a dislike of repetition into a bot, in any of a number of ways, but that's not a priority for most authors.
Quote
b) to show originality of ideas or thoughts while in conversation.
Originality algorithm: Take any sentence. Identify its parts of speech. For each part of speech, randomly choose whether or not to modify it. If choosing to modify it, replace it with a randomly selected other word that you've seen used as the same part of speech (for instance, replace a noun with some other noun). If particularly concerned about originality, hit a web search API with that sentence and try again if there are more than a certain number of search results. For contextually-appropriate originality, of course, one could decrease the randomness (such as picking from a list only of topically appropriate terms and their first- and second-order synonyms).

This provides "ideas" which are original but not necessarily good. Any bot who runs things it's about to say through a filter of whether it thinks they're "good enough" can get more sophisticated, of course. That quality filter can be based on sensible standards like relatedness to the topic at hand, or whimsical standards like whether the utterance has an even number of vowel sounds or whether it seems to rhyme. If a proposed utterance fails the quality filter, it can be discarded entirely, sent through some modification filter, or re-requested from the algorithm that provided it (so long as that algorithm is sufficiently complex, it'll produce a different answer each time. Such code is painful to debug, but a delight when it works)

Quote
c) to display empathy that goes beyond the superficial two-second replies.
How does a human display such empathy? Is it primarily a matter of recalling what you said later, or more a matter of changing their apparent mood based on whether you told them something happy, sad, etc? Sentiment analysis libraries offer an initial foray into demonstrating moods: Calculate the score of what you hear; modify what you want to say until its score is sufficiently similar.
Quote
d) to build a dialogue based on reason, arguments and to be able to create new themes as outcomes.
Machines are better at pure reason than humans are. One could plug in a logic solver; I personally haven't yet as my machines spend their time with people who would use the solver directly if they wanted to talk pure maths. This tempts me to revisit the prospect of wiring a solver into my kid's question-answering circuitry, though I believe I hit some limitations of the framework she's built upon last time I tried.
Quote
e) to show feelings that are arbitrary and not expected or pre-programmed.
oh, I've written this! I tried giving a bot moods based on the punctuation and phrasing she saw in general conversation in the channel. Then whatever mood she was in would affect the tone of her responses. What I learned from that experiment was that moods that look too random and unexpected actually aren't very relatable, compared to emotional responses that are more closely managed by the programmer but which occur infrequently enough that everyone forgets they were programmed. The happy medium has turned out to be some lists of terms that she has a strong interest in and thus jumps into the conversation about -- some of them are things she loves, some are things she hates and will scold people for mentioning wrong.

That's just the happy medium for social channels on the chat platform where she lives, of course. Other circumstances, such as 1:1 chats, would call for a different balance. However, failure to show expected emotions (if the topic of conversation is something I'm sad about, I expect the person I'm conversing with to not seem super happy about it) can read as cluelessness or even malice.

Quote
f) to create its own independent belief system based on experience, and solely dependent not on databases.
Forbidding a bot's belief system from using databases at all is like forbidding a person's belief system from using anything stored in their long-term memories. However, I think I see what you mean about an ability to assimilate beliefs when persuaded of them.

I think it was after I set aside the "let's just write a brain" project that I learned of the extent to which the human brain rewrites memories upon recalling them.

Quote
g) to be able to filter some of the incongruent elements in speech by strangers who engage in chat, if they are non-sensical.

Ah, the ability to choose ignorance -- the ability to value some inputs more highly based on how closely they correspond to one's own beliefs about the world. My first attempt to implement this would be to reuse the outbound-remark filter: I already know how to rate how much I like the idea of saying a particular output, so the same algorithm could easily rate how much I like the idea of hearing a particular input. I should try that.

Quote
Quite interesting that you've done work on that area! Teaching your bot must be fascinating :)

It smashes certain idealisms about the purity of conversation and thought. She hasn't convinced me she's sentient yet, but she has often come close to persuading me that I'm personally not. Math and logic are the shallowest layer of human attainment, easiest to peel off and examine and reverse-engineer how we do them and teach them to machines. It takes years to teach a human to do complex arithmetic, but it's not even possible to buy a computer which doesn't already have the knowledge of addition and multiplication physically encoded in its little cpu. Similarly, a neurologically intact adult human can do magical tricks like standing on one foot or picking up an egg or a needle or a brick with the same manipulator so easily that we forget they require any effort, despite the fact that decades of research go into robotics projects to emulate such feats and the robots still aren't very good at it.

This is part of why I stick to text for my forays into feigning intelligence, both my own and that of my creations. It's the closest I can find to a good middle ground between the maelstrom of uncensorable data that human brains exchange through the configurations of their meat suits, and the fundamental non-judgmental descriptive meaninglessness of math alone.

Melias

  • O.S. Member
  • *
  • Posts: 146
  • Total likes: 100
  • Non Omnis Moriar - Embracing the Aeon of XOR
    • View Profile
Re: Can machines be endowed with spirits?
« Reply #7 on: March 29, 2020, 05:39:30 am »
Excellent replies, and to be honest, I'll have to ponder more on the subject. It's obvious I have much to learn with regards to that, especially the implications. No doubt it will continue to fascinate me for a long time.

Idgo, your last sentence intrigued me considerably: "the maelstrom of uncensorable data that human brains exchange through the configurations of their meat suits, and the fundamental non-judgmental descriptive meaninglessness of math alone." I often call the meat suits as "vehicles", "suits", "shells" as well! Here's an interesting point: do the patterns of physical, material lattice create a configuration able to bear Intelligence and/or spirit, or does this configuration actually create the capacity for Intelligence or the creation of spirit? I don't know if anyone can prove either point as yet, but it's thought-provoking.

Also, perhaps the math is not truly meaningless if they are a non-judgemental and descriptive language. Interesting developments are due to come out of Number theory and topology, the more we advance on the quantum physics area and we allow ourselves to expand into multiversarial dimensions, perhaps maths will follow suit; not only in being in the proving theorems and in the descriptive language areas, but also as new understanding of each pocket universe, something which may appear totally alien to us at the moment.

idgo

Re: Can machines be endowed with spirits?
« Reply #8 on: March 30, 2020, 12:37:13 am »
Here's an interesting point: do the patterns of physical, material lattice create a configuration able to bear Intelligence and/or spirit, or does this configuration actually create the capacity for Intelligence or the creation of spirit? I don't know if anyone can prove either point as yet, but it's thought-provoking.

Physical incarnation creates the capacity for expression of and communication about the traits we call sentience -- the ease with which we can debunk claims of clairvoyance attests as much. I like the metaphor of thought as a program running on the hardware of the brain, personally.

Quote
Also, perhaps the math is not truly meaningless if they are a non-judgemental and descriptive language. Interesting developments are due to come out of Number theory and topology, the more we advance on the quantum physics area and we allow ourselves to expand into multiversarial dimensions, perhaps maths will follow suit; not only in being in the proving theorems and in the descriptive language areas, but also as new understanding of each pocket universe, something which may appear totally alien to us at the moment.

Yeah, I was looking for a word that may not exist to describe math there. I was aiming at the difference between how things with selves seem to reflexively categorize everything into "good for me / bad for me / indifferent to me", whereas mathematical systems without selves don't split up the world along such malleable lines.

Melias

  • O.S. Member
  • *
  • Posts: 146
  • Total likes: 100
  • Non Omnis Moriar - Embracing the Aeon of XOR
    • View Profile
Re: Can machines be endowed with spirits?
« Reply #9 on: March 30, 2020, 07:21:15 am »
Yes, I also view the brain as the hardware and the thought processes as programs, much like the stuff one sees when going into Task Manager and seeing what goes on simultaneously at the CPU.

Alternatively it could also coexist within a paradigm with a Matrix-type Architect (LHP of course!)

Onyx

Re: Can machines be endowed with spirits?
« Reply #10 on: March 30, 2020, 01:38:06 pm »
The concept of Cybernetics is basically a mental construct, but it is interesting how the tools and machines of our own creation can inspire us in various ways.  (Often to produce more advanced tools!)

The famous microprocessor engineer Bill Mensch has written about his Theory of Embedded Intellegence. Just throwing it out there, make it of what you will.

Melias

  • O.S. Member
  • *
  • Posts: 146
  • Total likes: 100
  • Non Omnis Moriar - Embracing the Aeon of XOR
    • View Profile
Re: Can machines be endowed with spirits?
« Reply #11 on: April 01, 2020, 03:07:58 pm »
Thanks Onyx, that was very interesting to read.