Keeping AIs honest

Casual conversation between friends. Anything goes (almost).
User avatar
Kim O'Hara
Former staff member
Posts: 7064
Joined: Fri Nov 16, 2012 1:09 am
Location: North Queensland, Australia

Re: Keeping AIs honest

Post by Kim O'Hara »

Karma Dondrup Tashi wrote: Thu Mar 09, 2023 12:36 am An irony? We have imposed the metaphor of an artificial mind on computers and then reimported the image of a thinking machine and imposed it upon our minds.
Yes, a lot of us have done both.

Metaphors are useful in helping us understand things but they become actively misleading when pushed too far. In this discussion I think we need to be particularly careful to recognise their limits.

:namaste:
Kim
User avatar
Agent Smith
Posts: 46
Joined: Wed Nov 23, 2022 4:46 pm

Re: Keeping AIs honest

Post by Agent Smith »

All of the above :good:

Change needs to be defined philosophically.
User avatar
PadmaVonSamba
Posts: 9443
Joined: Sat May 14, 2011 1:41 am

Re: Keeping AIs honest

Post by PadmaVonSamba »

I think that ‘artificial intelligence’ is a false term.

A more accurate term would be ‘synthesized taxis’.
taxis (pronounced tax-iss, not to be confused with yellow cars) is the ability of objects without brain function, usually plants, to respond and react accordingly to things in their environment.

For example, a growing cucumber plant sends out tendrils which seem to ‘search’ for things to latch onto and coil around. Some flowers move during the day to follow the Sun. Another example of living things which are able to respond to other objects outside of themselves are white blood cells which attack infections in the body, and sperm cells which swim to an ovum, “attracted” by chemicals it produces.

None of these things have what we recognize or define as “consciousness”. None of them even have brains.

In creating ‘Artificial Intelligence’ we have given computers the ability to perform taxis, to detect and conform to events outside of themselves. But if we call it “consciousness”, then we also have to say that plant cells and animal cells have consciousness, and we have to separate consciousness from taxis. What separates consciousness from taxis?

The question really comes down to self-awareness. Does a sperm “know” that it is swimming? Does a venus fly trap “know” that it is closing up on an insect? Do such actions manifest as an experience, thus, as an object of awareness to the plants and cells which perform these actions? If there is no experience, no ’conscious experience’ (meaning an experience that one is conscious of) then there is no consciousness. There is only very elaborate and complex mechanical function. All the events of taxis can be shown to be chemical reactions, or contractions and expansions creating movement, and so on. As long as the “how” of AI can be explained in terms of ‘ones and zeros’ then it is still the mere functioning of a machine.

The thing about the experience of awareness is that we don’t really know what it is. All we know is that it occurs. The Buddha says that there is no “self” that it occurs to, and he taught that sensory experiences of the subject arise along with their corresponding objects (the hearing of a dog barking arises when a dog is barking). So the problem in suggesting that a computer can possess the experience of awareness is in suggesting that “the machine possesses something which isn’t defined and has no characteristics of its own” —a suggestion that doesn’t lead us very far.
EMPTIFUL.
An inward outlook produces outward insight.
User avatar
Kim O'Hara
Former staff member
Posts: 7064
Joined: Fri Nov 16, 2012 1:09 am
Location: North Queensland, Australia

Re: Keeping AIs honest

Post by Kim O'Hara »

PadmaVonSamba wrote: Thu Mar 09, 2023 2:06 pm I think that ‘artificial intelligence’ is a false term.

A more accurate term would be ‘synthesized taxis’.
taxis (pronounced tax-iss, not to be confused with yellow cars) is the ability of objects without brain function, usually plants, to respond and react accordingly to things in their environment.

For example, a growing cucumber plant sends out tendrils which seem to ‘search’ for things to latch onto and coil around. Some flowers move during the day to follow the Sun. Another example of living things which are able to respond to other objects outside of themselves are white blood cells which attack infections in the body, and sperm cells which swim to an ovum, “attracted” by chemicals it produces.

None of these things have what we recognize or define as “consciousness”. None of them even have brains.

In creating ‘Artificial Intelligence’ we have given computers the ability to perform taxis, to detect and conform to events outside of themselves. But if we call it “consciousness”, then we also have to say that plant cells and animal cells have consciousness, and we have to separate consciousness from taxis. What separates consciousness from taxis?

The question really comes down to self-awareness. Does a sperm “know” that it is swimming? Does a venus fly trap “know” that it is closing up on an insect? Do such actions manifest as an experience, thus, as an object of awareness to the plants and cells which perform these actions? If there is no experience, no ’conscious experience’ (meaning an experience that one is conscious of) then there is no consciousness. There is only very elaborate and complex mechanical function. All the events of taxis can be shown to be chemical reactions, or contractions and expansions creating movement, and so on. As long as the “how” of AI can be explained in terms of ‘ones and zeros’ then it is still the mere functioning of a machine.
You had me worried for a moment there, but you set up a straw man only to pick it apart again and you ended up very close to where I was a few posts ago, at https://www.dharmawheel.net/viewtopic.p ... 16#p658016 so that's okay. :smile:
The thing about the experience of awareness is that we don’t really know what it is. All we know is that it occurs. The Buddha says that there is no “self” that it occurs to, and he taught that sensory experiences of the subject arise along with their corresponding objects (the hearing of a dog barking arises when a dog is barking). So the problem in suggesting that a computer can possess the experience of awareness is in suggesting that “the machine possesses something which isn’t defined and has no characteristics of its own” —a suggestion that doesn’t lead us very far.
Ultimately, yes. And we could speculate about whether an AI, if it existed, could become enlightened.
:thinking:
But I think the value of suggesting that a computer may acquire self-consciousness, and speculating about what might happen next, is more mundane. That is, we know what it feels like to be a person and to have free will, so it is possible to think about what may happen if a computer changes enough to likewise feel like a person and have free will.
And, since we really don't know what consciousness is or how it arises, we can't rule out the possibility of it happening,

:namaste:
Kim
MagnetSoulSP
Posts: 269
Joined: Fri Jun 30, 2023 1:45 am

Re: Keeping AIs honest

Post by MagnetSoulSP »

Lucky we are not programmed to feel suffering, since then how there could be any liberation! Oh, reprogram. Where is that button here now again?

Consciousness cannot be created by technology. But I often have been thinking, next life we will appear as a pc since we use this thing now so!

Since consciousness comes not from matter, not from a brain, I was reading this:
I'm pretty sure that we are "programmed" to feel suffering considering that we already do.

But to say consciousness can't be made by technology seems like an error in judgment. As far as the evidence seems to suggest consciousness is a emergent property with physical origins, though we aren't exactly sure how but the gap seems to be closing. I mean there are split brain patients with essentially two "people" in the same body.

But evidence seems to show consciousness comes from matter and I think we might get to the point of a sentient and conscious AI. Also I don't think the Dali Lama is the person to ask about this. Buddhism is a bit outdated when it comes to modern neuroscience. Even the notion of "Dependent arising" has been threatened by quantum physics.
Last edited by Ardha on Tue Mar 14, 2023 10:31 pm, edited 1 time in total.
User avatar
Johnny Dangerous
Global Moderator
Posts: 17092
Joined: Fri Nov 02, 2012 10:58 pm
Location: Olympia WA
Contact:

Re: Keeping AIs honest

Post by Johnny Dangerous »

Ardha wrote: Tue Mar 14, 2023 10:27 pm

But evidence seems to show consciousness comes from matter and I think we might get to the point of a sentient and conscious AI. Also I don't think the Dali Lama is the person to ask about this. Buddhism is a bit outdated when it comes to modern neuroscience. Even the notion of "Dependent arising" has been threatened by quantum physics.
Oh really? What evidence exactly do you have that consciousness is merely an emergent property of matter?

Where exactly is dependent arising contradicted by quantum physics?

Have you ever heard of the Mind and Life institute?

https://www.mindandlife.org/


As far as the evidence seems to suggest consciousness is a emergent property with physical origins, though we aren't exactly sure how but the gap seems to be closing. I mean there are split brain patients with essentially two "people" in the same body.
What? Split personality disorder is essentially a myth, what are you referring to?
https://diff.wikimedia.org/2017/09/14/w ... lit-brain/

If you are talking about this, it does not suggest that consciousness emerges from matter, only that there is a clear relationship between consciousness and the brain, senses, personality etc (duh), and that some functions of consciousness are dependent on a brain, which is not really controversial to anyone.

At this point in time science does not claim definitive proof wrt to the Hard Problem of Consciousness..usually what materialists do is what Daniel Dennet does, which is to claim it just isn't really a problem so it doesn't need to be explained. If anything, quantum physics seems to be leaning towards explanations of reality that acknowledge just how much of physical reality is dependent on an observer.

https://www.msn.com/en-us/news/technolo ... r-AA10JhQJ
Meditate upon Bodhicitta when afflicted by disease

Meditate upon Bodhicitta when sad

Meditate upon Bodhicitta when suffering occurs

Meditate upon Bodhicitta when you are scared

-Khunu Lama
User avatar
Konchog Thogme Jampa
Posts: 1175
Joined: Sun Oct 01, 2017 4:48 am
Location: Saha World/Hard to Take

Re: Keeping AIs honest

Post by Konchog Thogme Jampa »

Ardha wrote: Tue Mar 14, 2023 10:27 pm
Lucky we are not programmed to feel suffering, since then how there could be any liberation! Oh, reprogram. Where is that button here now again?

Consciousness cannot be created by technology. But I often have been thinking, next life we will appear as a pc since we use this thing now so!

Since consciousness comes not from matter, not from a brain, I was reading this:
I'm pretty sure that we are "programmed" to feel suffering considering that we already do.

But to say consciousness can't be made by technology seems like an error in judgment. As far as the evidence seems to suggest consciousness is a emergent property with physical origins, though we aren't exactly sure how but the gap seems to be closing. I mean there are split brain patients with essentially two "people" in the same body.

But evidence seems to show consciousness comes from matter and I think we might get to the point of a sentient and conscious AI. Also I don't think the Dali Lama is the person to ask about this. Buddhism is a bit outdated when it comes to modern neuroscience. Even the notion of "Dependent arising" has been threatened by quantum physics.
Ai is code often Python programmed by people
MagnetSoulSP
Posts: 269
Joined: Fri Jun 30, 2023 1:45 am

Re: Keeping AIs honest

Post by MagnetSoulSP »

Johnny Dangerous wrote: Wed Mar 15, 2023 1:08 am
Ardha wrote: Tue Mar 14, 2023 10:27 pm

But evidence seems to show consciousness comes from matter and I think we might get to the point of a sentient and conscious AI. Also I don't think the Dali Lama is the person to ask about this. Buddhism is a bit outdated when it comes to modern neuroscience. Even the notion of "Dependent arising" has been threatened by quantum physics.
Oh really? What evidence exactly do you have that consciousness is merely an emergent property of matter?

Where exactly is dependent arising contradicted by quantum physics?

Have you ever heard of the Mind and Life institute?

https://www.mindandlife.org/


As far as the evidence seems to suggest consciousness is a emergent property with physical origins, though we aren't exactly sure how but the gap seems to be closing. I mean there are split brain patients with essentially two "people" in the same body.
What? Split personality disorder is essentially a myth, what are you referring to?
https://diff.wikimedia.org/2017/09/14/w ... lit-brain/

If you are talking about this, it does not suggest that consciousness emerges from matter, only that there is a clear relationship between consciousness and the brain, senses, personality etc (duh), and that some functions of consciousness are dependent on a brain, which is not really controversial to anyone.

At this point in time science does not claim definitive proof wrt to the Hard Problem of Consciousness..usually what materialists do is what Daniel Dennet does, which is to claim it just isn't really a problem so it doesn't need to be explained. If anything, quantum physics seems to be leaning towards explanations of reality that acknowledge just how much of physical reality is dependent on an observer.

https://www.msn.com/en-us/news/technolo ... r-AA10JhQJ
For starters that we can turn consciousness on and off by stimulating an area of the brain. The other is that there is no evidence so far of a mind or consciousness without a body. There's more but that's what I know.

Mind and Life Institute sounds like BS from their webpage.

Split personality disorder is not a myth, not sure what you're getting at there.

As for that study it does heavily imply consciousness is tied to matter, considering impacting the brain does impact it. So much so that we are even able to create optical illusions by specifically stimulating certain areas of the brain. In short the evidence just keeps pointing to consciousness being tied to matter. You and others so far have yet to demonstrate evidence that substantiates this not being the case. Unless there is some sort of soul, or dualism is true (and neuroscience is indicating no) then you don't have much basis for that. My question though is why this seems to be an issue with spiritual types.

Last trying to invoke quantum physics without any math behind it is garbage. Not only does it misunderstand "observation" in quantum physics (it has nothing to do with observers, observation just means any interaction with a system) but there is no evidence it applies outside a closed quantum state. In short that link you gave is nonsense, the same when anyone tries to tie QM to consciousness. They ironically cite "schrodiners cat" which was used to illustrate the absurdity of applying quantum principles to macro level interactions. There's a lot more wrong in the link but it's essentially making the same mistaken of using interpretations (of which are really simplified views to makes sense of the math) to make claims about reality. Also the experiment cited is anything but solid and is still hotly debated. Again what the experiment in the article means largely depends on which interpretation of QM you are using.

Though usually any article talking about QM without mentioning the math is a red flag.

Pretty much if you see the terms quantum physics, objective reality, and consciousness together assume it's garbage for clicks (yes I have spoken to experts on this and they told me the same thing).

As to how it challenges Buddhism well there are plenty of uncaused causes at the quantum level with stuff popping into existence with no cause. But this is the quantum level. As someone told me QM is a measure of probability. The likelihood of it happening at the macro level is so small as to be essentially null.

PS: Though for your sake I would be hoping their interpretation and conclusion is not true as it would not only heavily imply solipsism but it would essentially nullify Buddhism too by extension.
Last edited by Ardha on Wed Mar 15, 2023 9:03 pm, edited 1 time in total.
User avatar
Johnny Dangerous
Global Moderator
Posts: 17092
Joined: Fri Nov 02, 2012 10:58 pm
Location: Olympia WA
Contact:

Re: Keeping AIs honest

Post by Johnny Dangerous »

Ardha wrote: For starters that we can turn consciousness on and off by stimulating an area of the brain. The other is that there is no evidence so far of a mind or consciousness without a body. There's more but that's what I know.
No, we can turn electrical signals in brains on and off, mess with neurotransmitters, map function of different areas, etc., then see reaction in organisms which we can record. That is not the same thing, it’s pure speculation that this causes a total cessation of the subjective experience of consciousness, or that consciousness is reducible to physical function only, and is non-falsifiable at any rate. Materialists tend to believe they simply don’t need to account for the fact that consciousness and qualia are non-falsifiable, but they are just avoiding the issue.
Mind and Life Institute sounds like BS from their webpage.
They’ve hosted and collaborated with all kinds of respected names in neuroscience, etc. So, here you just really don’t know what you are talking about.
Split personality disorder is not a myth, not sure what you're getting at there.
DID is real, two people in one body, not so much.
As for that study it does heavily imply consciousness is tied to matter, considering impacting the brain does impact it. So much so that we are even able to create optical illusions by specifically stimulating certain areas of the brain. In short the evidence just keeps pointing to consciousness being tied to matter. You and others so far have yet to demonstrate evidence that substantiates this not being the case. Unless there is some sort of soul, or dualism is true (and neuroscience is indicating no) then you don't have much basis for that. My question though is why this seems to be an issue with spiritual types.
Look up the Hard Problem of consciousness, it’s probably something you should know about if you want to run around claiming expertise here.

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5013033/
Pretty much if you see the terms quantum physics, objective reality, and consciousness together assume it's garbage for clicks (yes I have spoken to experts on this and they told me the same thing).
If they are like your Buddhism “experts” they aren’t worth listening to, but let us know what experts they are and what they said. For that matter, what is your own expertise in the area? Myself I know a bit of psychology due to education, but little on QM, I do note however that there has been a shift away from pure materialism and reductionism as regards the mind/body dichotomy in recent years, and I follow such debates as best I can as a layperson.

So, since you’re asserting yourself as an expert here evidently, what makes you one?
As to how it challenges Buddhism well there are plenty of uncaused causes at the quantum level with stuff popping into existence with no cause. But this is the quantum level. As someone told me QM is a measure of probability. The likelihood of it happening at the macro level is so small as to be essentially null.
“Someone told me”, sounds convincing.
PS: Though for your sake I would be hoping their interpretation and conclusion is not true as it would not only heavily imply solipsism but it would essentially nullify Buddhism too by extension.
I don’t think you understand Buddhism well enough to even know what would “nullify” it.
Meditate upon Bodhicitta when afflicted by disease

Meditate upon Bodhicitta when sad

Meditate upon Bodhicitta when suffering occurs

Meditate upon Bodhicitta when you are scared

-Khunu Lama
MagnetSoulSP
Posts: 269
Joined: Fri Jun 30, 2023 1:45 am

Re: Keeping AIs honest

Post by MagnetSoulSP »

No, we can turn electrical signals in brains on and off, mess with neurotransmitters, map function of different areas, etc., then see reaction in organisms which we can record. That is not the same thing, it’s pure speculation that this causes a total cessation of the subjective experience of consciousness, or that consciousness is reducible to physical function only, and is non-falsifiable at any rate. Materialists tend to believe they simply don’t need to account for the fact that consciousness and qualia are non-falsifiable, but they are just avoiding the issue.
Can and have. It sounds more like you're scared that it might be the case. Again, so far the evidence points to consciousness being emergent of the brain. Unless you're gonna appeal to consciousness as some uncaused and independent force like a "soul" or something (which would violate dependent arising) I'm not sure you're in any position to really dog on materialism that much. Seems to have a better track record than anything else so far. Your reaction isn't that different from Christians or other spiritual types any time science seems to threaten this aspect of being human. For some reason consciousness seems to be the last bastion of mysticism.
They’ve hosted and collaborated with all kinds of respected names in neuroscience, etc. So, here you just really don’t know what you are talking about.
Pretty sure I do since organizations like that tend to be highly prone to bias. Mixing religion and science tends to do that.
DID is real, two people in one body, not so much.
Then you don't know much about psychology. It is, effectively, two people in one body. Same with split brain patients.
Look up the Hard Problem of consciousness, it’s probably something you should know about if you want to run around claiming expertise here.

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5013033/
I've aware of the hard problem, but your link is 7 years old, that's more or less a fossil. Not to mention it doesn't really support it's own claims with that alternative framework. Though looking through it, it still supports the physicalist or materialist view of consciousness, though some parts in it raise red flags since it pretty much admits the alternate framework doesn't want to conclude it's physical, which is iffy. Again this just sounds like fear to me. Every time science seems to hint at material source of consciousness it seems to send spiritual types in a panic. I guess it's why AI is so threatening. Personally I find it exciting as it challenges our notions of our specialness.
I don’t think you understand Buddhism well enough to even know what would “nullify” it.
I know enough to know their conclusion would invalidate you. Not only would it call into question other sentient beings existing (since no objective reality if true) but also it would mean the idea of an independently existing "mind" or some odd that is the source of all this. It really would lead to solipsism.
“Someone told me”, sounds convincing.
I've talked to folks in the field and they pretty much say everyone is running rampant with interpretations when all they really are is the best guess of what the math means. But because they're weird people run off and try to use them for their own ends, like you and that article you cite. They also said any mention of QM without math isn't worth reading. The whole practice is heavy high level math that I KNOW you don't know and I don't either. But that won't stop people from dragging consciousness and other stuff to this.

The Wigner's friend I explained is not only highly contentious and debated, so far from settled (like a lot in QM) but it only applies to the quantum level. People forget that last part. But you can't really support the consciousness impacts quantum states without throwing out Buddhism with it.

I would recommend trying to find people who do this for their work and talking to them, but you'll likely get as far as I did as they said they would essentially have to teach me QM to explain it. This is why good science communication is important, otherwise people run off with things they don't understand. Just like the Alpha Wolf study.
User avatar
justsit
Posts: 1466
Joined: Wed Oct 21, 2009 9:24 pm
Location: Delaware

Re: Keeping AIs honest

Post by justsit »

Seems there are other applications of AI being implemented. The Law of Unintended Consequences appears to be coming into play....

https://www.theguardian.com/technology/ ... against-ai
User avatar
Johnny Dangerous
Global Moderator
Posts: 17092
Joined: Fri Nov 02, 2012 10:58 pm
Location: Olympia WA
Contact:

Re: Keeping AIs honest

Post by Johnny Dangerous »

Ardha wrote: Thu Mar 16, 2023 4:45 am
No, we can turn electrical signals in brains on and off, mess with neurotransmitters, map function of different areas, etc., then see reaction in organisms which we can record. That is not the same thing, it’s pure speculation that this causes a total cessation of the subjective experience of consciousness, or that consciousness is reducible to physical function only, and is non-falsifiable at any rate. Materialists tend to believe they simply don’t need to account for the fact that consciousness and qualia are non-falsifiable, but they are just avoiding the issue.
Can and have. It sounds more like you're scared that it might be the case. Again, so far the evidence points to consciousness being emergent of the brain. Unless you're gonna appeal to consciousness as some uncaused and independent force like a "soul" or something (which would violate dependent arising) I'm not sure you're in any position to really dog on materialism that much. Seems to have a better track record than anything else so far. Your reaction isn't that different from Christians or other spiritual types any time science seems to threaten this aspect of being human. For some reason consciousness seems to be the last bastion of mysticism.
They’ve hosted and collaborated with all kinds of respected names in neuroscience, etc. So, here you just really don’t know what you are talking about.
Pretty sure I do since organizations like that tend to be highly prone to bias. Mixing religion and science tends to do that.
DID is real, two people in one body, not so much.
Then you don't know much about psychology. It is, effectively, two people in one body. Same with split brain patients.
Look up the Hard Problem of consciousness, it’s probably something you should know about if you want to run around claiming expertise here.

https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5013033/
I've aware of the hard problem, but your link is 7 years old, that's more or less a fossil. Not to mention it doesn't really support it's own claims with that alternative framework. Though looking through it, it still supports the physicalist or materialist view of consciousness, though some parts in it raise red flags since it pretty much admits the alternate framework doesn't want to conclude it's physical, which is iffy. Again this just sounds like fear to me. Every time science seems to hint at material source of consciousness it seems to send spiritual types in a panic. I guess it's why AI is so threatening. Personally I find it exciting as it challenges our notions of our specialness.
I don’t think you understand Buddhism well enough to even know what would “nullify” it.
I know enough to know their conclusion would invalidate you. Not only would it call into question other sentient beings existing (since no objective reality if true) but also it would mean the idea of an independently existing "mind" or some odd that is the source of all this. It really would lead to solipsism.
“Someone told me”, sounds convincing.
I've talked to folks in the field and they pretty much say everyone is running rampant with interpretations when all they really are is the best guess of what the math means. But because they're weird people run off and try to use them for their own ends, like you and that article you cite. They also said any mention of QM without math isn't worth reading. The whole practice is heavy high level math that I KNOW you don't know and I don't either. But that won't stop people from dragging consciousness and other stuff to this.

The Wigner's friend I explained is not only highly contentious and debated, so far from settled (like a lot in QM) but it only applies to the quantum level. People forget that last part. But you can't really support the consciousness impacts quantum states without throwing out Buddhism with it.

I would recommend trying to find people who do this for their work and talking to them, but you'll likely get as far as I did as they said they would essentially have to teach me QM to explain it. This is why good science communication is important, otherwise people run off with things they don't understand. Just like the Alpha Wolf study.
So you won’t cite your “friend” and are still on a Buddhism site after two years basically…why again? Why are you here asking these questions if you aren’t interested in a Buddhist perspective?

There hasn’t been some new discovery in the past 7 years that would upend the Hard Problem, don’t troll me.
Meditate upon Bodhicitta when afflicted by disease

Meditate upon Bodhicitta when sad

Meditate upon Bodhicitta when suffering occurs

Meditate upon Bodhicitta when you are scared

-Khunu Lama
User avatar
Johnny Dangerous
Global Moderator
Posts: 17092
Joined: Fri Nov 02, 2012 10:58 pm
Location: Olympia WA
Contact:

Re: Keeping AIs honest

Post by Johnny Dangerous »

justsit wrote: Sun Mar 19, 2023 12:55 am Seems there are other applications of AI being implemented. The Law of Unintended Consequences appears to be coming into play....

https://www.theguardian.com/technology/ ... against-ai
So far these AIs do have a really uncanny ability to produce garbage art. I don’t doubt some good stuff can be created with the right people messing with it.
Meditate upon Bodhicitta when afflicted by disease

Meditate upon Bodhicitta when sad

Meditate upon Bodhicitta when suffering occurs

Meditate upon Bodhicitta when you are scared

-Khunu Lama
dharmafootsteps
Posts: 475
Joined: Sun Apr 30, 2017 8:57 am

Re: Keeping AIs honest

Post by dharmafootsteps »

Johnny Dangerous wrote: Sun Mar 19, 2023 2:08 am
justsit wrote: Sun Mar 19, 2023 12:55 am Seems there are other applications of AI being implemented. The Law of Unintended Consequences appears to be coming into play....

https://www.theguardian.com/technology/ ... against-ai
So far these AIs do have a really uncanny ability to produce garbage art. I don’t doubt some good stuff can be created with the right people messing with it.
Perhaps you haven't kept up to date with the rate of progress. Is it time to start hanging AI art in the Louvre? No. But these things are probably already better than the average human artist. That's the trend in many tasks and fields just now, they're not better than experts, but they can perform closer to the top end of the curve than the bottom.
User avatar
Kim O'Hara
Former staff member
Posts: 7064
Joined: Fri Nov 16, 2012 1:09 am
Location: North Queensland, Australia

Re: Keeping AIs honest

Post by Kim O'Hara »

dharmafootsteps wrote: Sun Mar 19, 2023 9:31 am
Johnny Dangerous wrote: Sun Mar 19, 2023 2:08 am
justsit wrote: Sun Mar 19, 2023 12:55 am Seems there are other applications of AI being implemented. The Law of Unintended Consequences appears to be coming into play....

https://www.theguardian.com/technology/ ... against-ai
So far these AIs do have a really uncanny ability to produce garbage art. I don’t doubt some good stuff can be created with the right people messing with it.
Perhaps you haven't kept up to date with the rate of progress. Is it time to start hanging AI art in the Louvre? No. But these things are probably already better than the average human artist. That's the trend in many tasks and fields just now, they're not better than experts, but they can perform closer to the top end of the curve than the bottom.
Can you share examples?

:popcorn:
Kim
dharmafootsteps
Posts: 475
Joined: Sun Apr 30, 2017 8:57 am

Re: Keeping AIs honest

Post by dharmafootsteps »

Kim O'Hara wrote: Sun Mar 19, 2023 11:34 am
dharmafootsteps wrote: Sun Mar 19, 2023 9:31 am
Johnny Dangerous wrote: Sun Mar 19, 2023 2:08 am

So far these AIs do have a really uncanny ability to produce garbage art. I don’t doubt some good stuff can be created with the right people messing with it.
Perhaps you haven't kept up to date with the rate of progress. Is it time to start hanging AI art in the Louvre? No. But these things are probably already better than the average human artist. That's the trend in many tasks and fields just now, they're not better than experts, but they can perform closer to the top end of the curve than the bottom.
Can you share examples?

:popcorn:
Kim
I haven't done any scouting to find the best examples, but here are a few tweets that pop up just now when I search for 'Midjourney v5', which is one of the latest models (they are threads, so if you click "Read the full conversation on Twitter" they have many more images).




User avatar
Kim O'Hara
Former staff member
Posts: 7064
Joined: Fri Nov 16, 2012 1:09 am
Location: North Queensland, Australia

Re: Keeping AIs honest

Post by Kim O'Hara »

:thanks:
Kim
dharmafootsteps
Posts: 475
Joined: Sun Apr 30, 2017 8:57 am

Re: Keeping AIs honest

Post by dharmafootsteps »

I think there are a couple of key points worth remembering when evaluating these things.

1) The pace of progress. Trying to judge the technology as a whole based on their exact capacities in specific tasks right now is pointless. All these models are improving leaps and bounds every few months, weeks in some cases.

2) What they're actually being trained for. The image generators are not being trained for artistic talent, but to accurately follow a text prompt. If they produce an image that fulfils the instructions provided, that's a success. And the language models are being trained to "understand" language, not to have subject-level expertise. People are often unimpressed when a model doesn't have a perfect answer in their niche subject of interest, as in the "AI Dzogchen" thread here, but success in that case is giving an appearance of understanding the question and giving a coherent answer, not being equivalent to a Dzogchen teacher in its nuanced understanding of Dzogchen.

These things are eminently trainable, so more specific implementations will come e.g. image generators trained for a specific type of artistic taste, or language models built on top of a specific knowledge base. For example you could already create a bot that answers Buddhist questions with more depth and breadth than all but a few of the most educated and experienced posters here, simply by giving it a corpus of Dharma texts on the backend.
User avatar
Aryjna
Posts: 1626
Joined: Mon Mar 27, 2017 12:45 pm

Re: Keeping AIs honest

Post by Aryjna »

dharmafootsteps wrote: Sun Mar 19, 2023 1:04 pm For example you could already create a bot that answers Buddhist questions with more depth and breadth than all but a few of the most educated and experienced posters here, simply by giving it a corpus of Dharma texts on the backend.
I don't think this is really true. First, there is the serious problem of translations of dharma texts using all kinds of different terminology and using it in different ways. In addition, the meaning of many translations may very well be wrong, given that translation requires the correct interpretation of the material, something of which not all translators are capable. Also, AI would be trained by (or the end result of the training would be evaluated by) people who lack the expertise of the most educated and experienced posters here, and as a result it would only be trained to "fool" the people who took part in its training by sounding as it if it is responding with that breadth and depth but not to actually answer with that breadth and depth. This makes it all the more dangerous/misleading, because it sounds very believable to those who don't know better.
dharmafootsteps
Posts: 475
Joined: Sun Apr 30, 2017 8:57 am

Re: Keeping AIs honest

Post by dharmafootsteps »

Aryjna wrote: Sun Mar 19, 2023 1:23 pm
dharmafootsteps wrote: Sun Mar 19, 2023 1:04 pm For example you could already create a bot that answers Buddhist questions with more depth and breadth than all but a few of the most educated and experienced posters here, simply by giving it a corpus of Dharma texts on the backend.
I don't think this is really true.
I agree with most of what you said, but I don't think it really affects my point, so I'll stand by it
Aryjna wrote: Sun Mar 19, 2023 1:23 pm First, there is the serious problem of translations of dharma texts using all kinds of different terminology and using it in different ways. In addition, the meaning of many translations may very well be wrong, given that translation requires the correct interpretation of the material, something of which not all translators are capable.
Agreed, but this affects anyone trying to learn Dharma, it's not specific to AI. So I don't think it's of consequence in whether a bot could answer questions better than the average forum participant. However, it would be one of the things that separates the most knowledgable posters from the bot (and from everyone else).
Aryjna wrote: Sun Mar 19, 2023 1:23 pm Also, AI would be trained by (or the end result of the training would be evaluated by) people who lack the expertise of the most educated and experienced posters here, and as a result it would only be trained to "fool" the people who took part in its training by sounding as it if it is responding with that breadth and depth but not to actually answer with that breadth and depth.
That's generally not how something like this would be accomplished, it is possible to do human-in-the-loop training, but it's not what I was referring to here. The OpenAI API allows you to give a model "embeddings" which provide context for it in constructing an answer. So what I was referring to would be to give the model a broad selection of texts to be sourced from when answering. Essentially, rather than trying to come up with an answer based on everything it's ever been exposed to, it comes up with an answer based on a selection of high quality material. Yes, what one might consider high quality is subjective, but again that's not an AI specific problem.
Aryjna wrote: Sun Mar 19, 2023 1:23 pm This makes it all the more dangerous/misleading, because it sounds very believable to those who don't know better.
I'm definitely not implying it should be listened to as authoritative. Personally there are only about 4 or 5 people here whose posts I consider to have some degree of authority, and most of them are Dharma teachers themselves. Most of the rest (including myself) are not trained to a level that their posts should be treated with anything other than a large pinch of salt. How you consider the average poster here is obviously subjective, but basically, I'm not suggesting you could make a bot that replaces a teacher, or a highly educated/experienced practitioner, just that it could be of a pretty good standard compared to the average Buddhist forum participant.

Anyway, exactly how good it might or might not be compared to the average poster wasn't the point I was trying to make. It was that the base models people are being exposed to at the moment are not being made to be subject-level experts, and evaluating them on those terms will cause one to underestimate them. More specific implementations will be made for specific use cases, and are even already possible given current technology.
Last edited by dharmafootsteps on Sun Mar 19, 2023 2:06 pm, edited 3 times in total.
Post Reply

Return to “Lounge”