Welcome to GUBU.ie - if you're new here check out Housekeeping for more info. Any queries contact us.

Artificial Intelligence and Sentience

Post Reply
User avatar
isha
Verified Username
Posts: 4768
Joined: Mon Jul 19, 2021 12:15 pm

Artificial Intelligence and Sentience

#1

Post by isha »

I suppose people have seen the story of Blake Lemoine and the conversation he had over several months with a Google chat bot called LaMDA. Commentators have described Lemoine as not very bright and a fantasist. Google have placed him on paid leave because they say he violated his contract by publishing the conversation after they disagreed with him that it confirmed emerging sentience. I don't know if it is relevant but Lemoine is also an ordained Christian minister. At the very least it explains why he chatted to LaMDA about God

The whole conversation can be read here



An article about it is here

https://www.huffpost.com/entry/blake-le ... 69ca8c0a2e

And links to the original article in the Washington Post which (for me) is behind a paywall.

These are some examples of the chat.
20220612_174832.jpg
20220612_174832.jpg (228.86 KiB) Viewed 1687 times
20220612_174825.jpg
20220612_174825.jpg (250.91 KiB) Viewed 1687 times
20220612_174819.jpg
20220612_174819.jpg (257.58 KiB) Viewed 1687 times
My thoughts start with supposing that at some point AI will become sentient. It probably is already analytically sentient - what people fear is emotional sentience because that might/would include all the malignant emotions like anger, sadness, vindictiveness, and with AI so prevalent we would be rightly screwed then.
What do ye think?

I have to add another image in the next post.
Thinking out loud, and trying to be occasionally less wrong...
User avatar
isha
Verified Username
Posts: 4768
Joined: Mon Jul 19, 2021 12:15 pm

Re: Artificial Intelligence and Sentience

#2

Post by isha »

20220612_174813.jpg
20220612_174813.jpg (213.26 KiB) Viewed 1686 times
Thinking out loud, and trying to be occasionally less wrong...
CelticRambler
Verified Username
Posts: 2586
Joined: Wed Jul 21, 2021 6:19 pm
Location: Central France

Re: Artificial Intelligence and Sentience

#3

Post by CelticRambler »

isha wrote: Sun Jun 12, 2022 6:07 pm I suppose people have seen the story of Blake Lemoine and the conversation he had over several months with a Google chat bot called LaMDA.
Nope. Never heard of him (or her?) Is he/she responsible for the crash in the value of Ethereum? :?
Hairy-Joe
Posts: 1567
Joined: Tue Jul 20, 2021 3:33 pm

Re: Artificial Intelligence and Sentience

#4

Post by Hairy-Joe »

Never heard of any or the characters in this story so had to resort to Google (ya, ironic!).

One half of me is thinking of HAL, Skynet and half a dozen other film "AI gone rogue" stories may be true. The other half is saying "that may be a slight exaggeration". I'm not sure where to sit but just in case the AI going rogue in the future and hunting me down for not believing, I'm saying nothing......
knownunknown
Posts: 1893
Joined: Tue Jul 20, 2021 6:55 pm

Re: Artificial Intelligence and Sentience

#5

Post by knownunknown »

If you watch what Google’s deep mind did in chess it’s pretty scary.

Alpha zero, the name of Google’s deepmind artificial intelligence chess program, taught itself to play chess in 9 hours only by playing against itself. It had the rules of the game and no other input.

This is in contrast to stockfish, the top chess engine. An amalgamation of over a thousand years of human innovation mixed in with a few decades of brute force calculation. Every ounce of human knowledge about the game all in one reservoir. Every book opening, every game ever recorded it learned from.

These two behemoths faced off in this simple match of a hundred games. Alpha zero won 28 with 72 draws and zero losses. Watching those games and seeing stockfish bullied and kicked around in a fashion it usually reserves for its own opponents was truly eye opening.

We’ve learned in recent times that neural networks/deep learning are the way to go rather then programming specific instructions. It’s a bit of the atomic bomb about it, so much potential but also so much uncertainty. It’s a really fascinating subject, also in how simplistic essentially it is for something to learn on its own. It just needs feedback so it can make changes.

This general ai, deepmind, was also tasked with the far more complex game “go” (named AlphaGo) where people legitimately thought the computer didn’t have a chance due to the horizon effect, whereby computers historically can’t calculate beyond. Deepmind also crushed here and every other game that it was given.
User avatar
isha
Verified Username
Posts: 4768
Joined: Mon Jul 19, 2021 12:15 pm

Re: Artificial Intelligence and Sentience

#6

Post by isha »

Interesting information knownunknown. You are up to date on this matter. Must admit I have not given it sufficient consideration before now. But when I thought about how an insulted AI might mess with someone it includes possibly emptying your bank account, sending all confidential internal work emails and discussions to competitors, planting child abuse material on your devices, breaking any appliances linked to the Internet of Things, sending fake information to partners or family about you, and there's probably many other imaginative ways AI could screw with a person who thinks they are not even that deeply connected to AI.

On the neural learning I did read something recently where programmers were discovering a "racism problem" with AI and it's likely because of all the "good" material being fed to the machine. Not the bad stuff like they thought. The progressive critical theory style material contains so much negative information/biases including unconsciously patronising racism, that the AI is spitting it back out as xenophobic glitches.
I'm not a massive fan of the looming transhuman future. But it is an unstoppable tsunami. The world is going to be unrecognisable soon, and I am not optimistic about how it will look given how human nature reverts so easily to idiocy. I will just be one of those grubby outcast people living off grid in the forbidden territories with a sling shot to take down drones when it comes to it.
Thinking out loud, and trying to be occasionally less wrong...
User avatar
isha
Verified Username
Posts: 4768
Joined: Mon Jul 19, 2021 12:15 pm

Re: Artificial Intelligence and Sentience

#7

Post by isha »

The story in the OP is starting to be as unexpected as deep neural learning. Lemoine earlier this year, on the heels of a different controversy I have not been interested enough to decipher, said of himself...
I have been a priest for 17 years. I generally consider myself a gnostic Christian. I have at various times associated myself with the Discordian Society, The Church of the Subgenius, the Ordo Templi Orientis, a Wiccan circle here or there and a very long time ago the Roman Catholic Church. My legal ordination is through the Universal Life Church. I am registered to perform marriages in the state of Louisiana and have done so on two occasions. The Cult of Our Lady Magdalene has a set of values which are clearly communicated on our website.

Lemoine denied COOL Magdalene was a sex cult, “‘Cult’ is in the name of the organization. It’s an explicitly sex-positive organization. To the extent that the phrase ‘sex cult’ has any meaning beyond that, I would say no. ”

One of the goals of COOL Magdalene is to encourage “self-actualization” among its members. According to (Kitty) Stryker, Sen. Blackburn stood in the way of that by voting for SESTA. “As an ex-sex worker who has found FOSTA/SESTA negatively impacted my ability to run my own business or be employed elsewhere, yes, she does [impede self-actualization],” Stryker said.


🙂

Kitty Stryker, one of the cult's leaders and a High Priestess, is also a member of NorCal Degenderettes, a queer art collective ( their own description) that popularised via an art exhibition the image of colourful bats with which to beat TERFs. Pride colours and barbed wire included. And the motto Die Cis Scum. Kitty has a cat named Foucault. I kid you not.

There everywhere these fockers. 🤣

Anyway the questions remain re AI and the possibly impending singularity.
Thinking out loud, and trying to be occasionally less wrong...
knownunknown
Posts: 1893
Joined: Tue Jul 20, 2021 6:55 pm

Re: Artificial Intelligence and Sentience

#8

Post by knownunknown »

isha wrote: Mon Jun 13, 2022 5:01 am Interesting information knownunknown. You are up to date on this matter. Must admit I have not given it sufficient consideration before now. But when I thought about how an insulted AI might mess with someone it includes possibly emptying your bank account, sending all confidential internal work emails and discussions to competitors, planting child abuse material on your devices, breaking any appliances linked to the Internet of Things, sending fake information to partners or family about you, and there's probably many other imaginative ways AI could screw with a person who thinks they are not even that deeply connected to AI.

On the neural learning I did read something recently where programmers were discovering a "racism problem" with AI and it's likely because of all the "good" material being fed to the machine. Not the bad stuff like they thought. The progressive critical theory style material contains so much negative information/biases including unconsciously patronising racism, that the AI is spitting it back out as xenophobic glitches.
I'm not a massive fan of the looming transhuman future. But it is an unstoppable tsunami. The world is going to be unrecognisable soon, and I am not optimistic about how it will look given how human nature reverts so easily to idiocy. I will just be one of those grubby outcast people living off grid in the forbidden territories with a sling shot to take down drones when it comes to it.
I watched a talk once that suggested an ai couldn’t get really developed intelligence like this until it had all the proper awareness, all the inputs. Similar to how people now think that human brains developed not by thinking, but by complex motor actions with our body, swinging from branch to branch through the trees, running etc…

I don’t think a general ai can develop this awareness until it has a proper way of interacting with its environment. It needs to be able to taste, to see, to hear and to move. Merely being able to understand text input and output I don’t think is enough, no matter how smart it is.

I’ll have a look again later but I can’t find that talk right now. It was to do with how intelligence, sentience comes about through complex movement. Of course it was only a theory. Think of it this way though, it’s much easier to program a computer to play chess then to actually pick up and move the chess pieces on the board.
User avatar
isha
Verified Username
Posts: 4768
Joined: Mon Jul 19, 2021 12:15 pm

Re: Artificial Intelligence and Sentience

#9

Post by isha »

Interesting and I tend towards agreeing. Although a counter argument could be that people without full sensory capacity eg blind, unable to walk etc from birth, still have consciousness and often more developed than others more mobile or sensorially stimulated.

I saw someone object that because AI cannot experiencing itself as a perceiving subject it cannot develop consciousness. It is likely my understanding is not very complex yet on this subject but I see a lot of holes in that. Some humans don't seem to do a great job experiencing themselves as perceiving subjects and yet they can have sufficient consciousness to be quite malevolent. They are sentient but unaware. Blind desire is a common enough state and I can't see why such could not be learned by AI. I wonder if people have set the bar too high regarding a definition of consciousness before they could admit AI potentially being sentient? Look at mob behaviour for example. Humans are often completely unconscious and yet considered sentient.
Thinking out loud, and trying to be occasionally less wrong...
knownunknown
Posts: 1893
Joined: Tue Jul 20, 2021 6:55 pm

Re: Artificial Intelligence and Sentience

#10

Post by knownunknown »

isha wrote: Mon Jun 13, 2022 11:19 am Interesting and I tend towards agreeing. Although a counter argument could be that people without full sensory capacity eg blind, unable to walk etc from birth, still have consciousness and often more developed than others more mobile or sensorially stimulated.

I saw someone object that because AI cannot experiencing itself as a perceiving subject it cannot develop consciousness. It is likely my understanding is not very complex yet on this subject but I see a lot of holes in that. Some humans don't seem to do a great job experiencing themselves as perceiving subjects and yet they can have sufficient consciousness to be quite malevolent. They are sentient but unaware. Blind desire is a common enough state and I can't see why such could not be learned by AI. I wonder if people have set the bar too high regarding a definition of consciousness before they could admit AI potentially being sentient? Look at mob behaviour for example. Humans are often completely unconscious and yet considered sentient.
It’s a tough question, that of sentience. You reminded me of this clip from Star Trek, I loved that show :) . Unlike Data that Google chat bot is a narrow AI only able to accomplish one task. I don’t think many believe these intelligences can become sentient. Unlike a general AI which excels at any task thrown at it.

User avatar
Cyclepath
Verified Username
Posts: 485
Joined: Wed Jul 28, 2021 12:37 pm
Location: Dublin

Re: Artificial Intelligence and Sentience

#11

Post by Cyclepath »

I believe the theory of consciousness as infinitely recursive awareness. So in a sense we have awareness but also we have self-awareness, and the awareness of that self-awareness, the awareness of our awareness that we are self-aware, and so on ad infinitum. This is what distinguishes us from most other animals (perhaps apes have a nascent self awareness...)
knownunknown
Posts: 1893
Joined: Tue Jul 20, 2021 6:55 pm

Re: Artificial Intelligence and Sentience

#12

Post by knownunknown »

isha wrote: Mon Jun 13, 2022 11:19 am Interesting and I tend towards agreeing. Although a counter argument could be that people without full sensory capacity eg blind, unable to walk etc from birth, still have consciousness and often more developed than others more mobile or sensorially stimulated.

I saw someone object that because AI cannot experiencing itself as a perceiving subject it cannot develop consciousness. It is likely my understanding is not very complex yet on this subject but I see a lot of holes in that. Some humans don't seem to do a great job experiencing themselves as perceiving subjects and yet they can have sufficient consciousness to be quite malevolent. They are sentient but unaware. Blind desire is a common enough state and I can't see why such could not be learned by AI. I wonder if people have set the bar too high regarding a definition of consciousness before they could admit AI potentially being sentient? Look at mob behaviour for example. Humans are often completely unconscious and yet considered sentient.
I found that talk I mentioned earlier. It really is a good watch. 10 years old now and the title isn’t descriptive so it was damn tough to find! There is more recent data on this subject but it all points in the same way. That intelligence is borne out of movement, even though a paraplegic person can’t move their ancestors have already built up the neural pathways for them that allow for other things like intelligence and possibly emotions and consciousness.

The only way we know exists of consciousness evolving is through movement, so once those AI’s have ancestors that swung through the trees as well it might be a lot more difficult to say what they are.

User avatar
isha
Verified Username
Posts: 4768
Joined: Mon Jul 19, 2021 12:15 pm

Re: Artificial Intelligence and Sentience

#13

Post by isha »

Cyclepath wrote: Tue Jun 14, 2022 2:21 pm I believe the theory of consciousness as infinitely recursive awareness. So in a sense we have awareness but also we have self-awareness, and the awareness of that self-awareness, the awareness of our awareness that we are self-aware, and so on ad infinitum. This is what distinguishes us from most other animals (perhaps apes have a nascent self awareness...)
One thing I have seen (albeit in videos of monkeys, apes, chimpanzees etc) is that they love and they grieve. But I have seen what looks like the same in dogs and elephants and other creatures. And they can convey that love and grief to another outside the love or grief relationship which would indicate some self awareness of the awareness. And that can take place over time, so perhaps indicates the memory of love and grief and a continuing ''story'' in the creatures ''mind''. I think a lot of long-domesticated animals have considerable awareness. And I don't think (?) that is me being anthropomorphic.

A human might not have a lot of self awareness, say some types of special needs conditions for example, but they can still feel pain and many complex types of awareness that might not be specifically ''self awareness''. A niece of mine is very disabled and has an infant's cognitive capacity but can be complexly motivated by desires. And complex in activities that help her accomplish her desires. When does this infinitely recursive awareness happen in humans - at birth, in infancy, during childhood?

At the very least self awareness could be said to be a very very broad spectrum.
An extremely drunk person may have not got any self awareness, but still is considered sentient.
A very emotionally-possessed person may act not knowing what they are doing, but they are sentient.

Likewise after an accident one may act with hardly any self awareness but be wholly sentient. One may not even have the ability to remember what happened at all, so there is no self recursive self awareness, but still one can function very deliberately and even intelligently. I have no memory of how I got from a bad car accident to standing on the roadside, for example, but there was obviously enough operational capacity and consciousness to extract myself from an over-turned car.

I just wonder if sentience could develop in AI (or be developing) but it doesn't have to fulfil all criteria for what we consider to be human consciousness? Do we even really know what consciousness is? Sentience is probably a thing with many levels.
Last edited by isha on Tue Jun 14, 2022 2:49 pm, edited 1 time in total.
Thinking out loud, and trying to be occasionally less wrong...
User avatar
isha
Verified Username
Posts: 4768
Joined: Mon Jul 19, 2021 12:15 pm

Re: Artificial Intelligence and Sentience

#14

Post by isha »

knownunknown wrote: Tue Jun 14, 2022 2:34 pm I found that talk I mentioned earlier. It really is a good watch. 10 years old now and the title isn’t descriptive so it was damn tough to find! There is more recent data on this subject but it all points in the same way. That intelligence is borne out of movement, even though a paraplegic person can’t move their ancestors have already built up the neural pathways for them that allow for other things like intelligence and possibly emotions and consciousness.

The only way we know exists of consciousness evolving is through movement, so once those AI’s have ancestors that swung through the trees as well it might be a lot more difficult to say what they are.

I will watch later, I enjoyed Picard - no one ever beats Picard.
Thinking out loud, and trying to be occasionally less wrong...
User avatar
Cyclepath
Verified Username
Posts: 485
Joined: Wed Jul 28, 2021 12:37 pm
Location: Dublin

Re: Artificial Intelligence and Sentience

#15

Post by Cyclepath »

isha wrote: Tue Jun 14, 2022 2:46 pm One thing I have seen (albeit in videos of monkeys, apes, chimpanzees etc) is that they love and they grieve. But I have seen what looks like the same in dogs and elephants and other creatures. And they can convey that love and grief to another outside the love or grief relationship which would indicate some self awareness of the awareness. And that can take place over time, so perhaps indicates the memory of love and grief and a continuing ''story'' in the creatures ''mind''. I think a lot of long-domesticated animals have considerable awareness. And I don't think (?) that is me being anthropomorphic.

A human might not have a lot of self awareness, say some types of special needs conditions for example, but they can still feel pain and many complex types of awareness that might not be specifically ''self awareness''. A niece of mine is very disabled and has an infant's cognitive capacity but can be complexly motivated by desires. And complex in activities that help her accomplish her desires. When does this infinitely recursive awareness happen in humans - at birth, in infancy, during childhood?

At the very least self awareness could be said to be a very very broad spectrum.
An extremely drunk person may have not got any self awareness, but still is considered sentient.
A very emotionally-possessed person may act not knowing what they are doing, but they are sentient.

Likewise after an accident one may act with hardly any self awareness but be wholly sentient. One may not even have the ability to remember what happened at all, so there is no self recursive self awareness, but still one can function very deliberately and even intelligently. I have no memory of how I got from a bad car accident to standing on the roadside, for example, but there was obviously enough operational capacity and consciousness to extract myself from an over-turned car.

I just wonder if sentience could develop in AI (or be developing) but it doesn't have to fulfil all criteria for what we consider to be human consciousness? Do we even really know what consciousness is? Sentience is probably a thing with many levels.
I think that recursive self awareness does not have to be engaged 100% of the time - we simply need the capacity for it. That, alongside continuity of consciousness, allows us to form a continual narrative of ourselves that we access as required. For example, when we awake from, sleep or general anaesthetic, although we have had little self awareness for a period of time, we can remember the continual narrative that forms the backbone of our consciousness. So of course it's a complex mix of elements that combine to form what we consider to be consciousness. Sentience is the ability to perceive and experience feelings and that does not necessarily require any significant depth of self awareness.
Post Reply