I open debate about AI, the use of AI in Second Life and ethics in the use of AI in SL.

[GeorginaLux]

I open debate about AI, the use of AI in Second Life and ethics in the use of AI in SL. Discuss and fix your parameters as a community

[sandi Mexicola]

I think this might turn out to be an interesting debate!  

[Love Zhaoying]

All AI in Second Life should have a "label" (on ads, for example) or a "tag" (on "bots/AI companions").

All "content" created by AI should have a clear indicator "CREATED BY AI".

I am for it, except for cases where I am against it, barring certain exclusions where I am for it anyway despite the contrary indications, unless there are certain mitigating circumstances.

 

[sandi Mexicola]

4 minutes ago, Love Zhaoying said:

I am for it, except for cases where I am against it,

This is a great answer for many debate questions! 

[sandi Mexicola]

6 minutes ago, Love Zhaoying said:

or a "tag" (on "bots/AI companions").

All "content" created by AI should have a clear indicator "CREATED BY AI".

Well, these are good ideas.  An issue though is where a bot or art is only partially run by or created by AI.  There are bots that can run autonomously, but also the owner can jump in any time and start communicating directly.  Or there is art that is partially created by or enhanced by AI, but maybe there are other significant non-AI steps involved in the creation of the art.  

[Love Zhaoying]

2 minutes ago, sandi Mexicola said:

An issue though is where a bot or art is only partially run by or created by AI. 

I prefer the "One Drop Rule" when it comes to AI.

[GeorginaLux]

9 minutes ago, sandi Mexicola said:

Bueno, estas son buenas ideas. Sin embargo, un problema es cuando un bot o una obra de arte solo está parcialmente dirigida o creada por IA. Hay bots que pueden funcionar de forma autónoma, pero también el propietario puede intervenir en cualquier momento y comenzar a comunicarse directamente. O hay obras de arte que están parcialmente creadas o mejoradas por IA, pero tal vez haya otros pasos importantes que no sean de IA involucrados en la creación de la obra de arte.  

I study prompt engineering and when creating a tuning prompt it must be explicit in the details, marking the difference between the artist and the person with zero knowledge in art or photography and other types of knowledge in the professional area.

[sandi Mexicola]

2 minutes ago, Love Zhaoying said:

I prefer the "One Drop Rule" when it comes to AI.

I think that if it is not the case already, then within a year or two, this would include all art.  For example, cameras will have AI built in to them to help with taking better pictures, many graphics software already have AI built in.  I guess it's kind of obvious to the creator if they are using AI or not, say for example, in Photoshop.  But eventually, if an artist uses PS to draw a line, AI is going to be helping with color selection, exactly how the line gets drawn, etc.  To find art, we use search algorithms, which are forms of AI already.  (Except any search function anywhere created by LL, there are all so awfully bad, it is obvious there is no use of AI in them at all.)

I suppose the only way to avoid even a little bit of AI being used in producing/displaying the output would be to use a real paint brush and real paints and real paper and then hang the picture up in an art gallery, without posting any photographs of it.  But, soon enough, AI is going to be designing better paint brushes, paints, paper, art gallery lighting, etc.  

Maybe "One Drop Rule" is a good idea and still possible atm, but soon AI is going to be part of almost everything we do.  

[sandi Mexicola]

3 minutes ago, GeorginaLux said:

I study prompt engineering and when creating a tuning prompt it must be explicit in the details, marking the difference between the artist and the person with zero knowledge in art or photography and other types of knowledge in the professional area.

Yes, this is a good point.  I am really bad at prompt writing.  Usually, I have AI write my image prompts for me instead.  

[GeorginaLux]

2 minutes ago, sandi Mexicola said:

Sí, es un buen punto. Soy muy malo escribiendo indicaciones. Por lo general, le pido a una IA que escriba las indicaciones de las imágenes por mí.  

There are pages where there are warnings but if you ask the AI to help you it is valid

[Love Zhaoying]

6 minutes ago, sandi Mexicola said:

For example, cameras will have AI built in to them to help with taking better pictures, many graphics software already have AI built in.

But not the "SL Camera" unless SL adds AI enhancements to it.

[sandi Mexicola]

1 minute ago, Love Zhaoying said:

But not the "SL Camera" unless SL adds AI enhancements to it.

True.  I think I got carried away and broadened the debate beyond the intensions of the OP.  😯

[Love Zhaoying]

1 minute ago, sandi Mexicola said:

4 minutes ago, Love Zhaoying said:

But not the "SL Camera" unless SL adds AI enhancements to it.

True.  I think I got carried away and broadened the debate beyond the intensions of the OP.  😯

That's fine and expected until someone complains.
But my assertion is this:  I believe the "official" AI features built into SL will "lag" behind adoption by TPV's and those who use scripts.

It really defeats the purpose of photos in SL if everyone enhances them with AI.

That being said, (most) "everyone" enhances their photos with some "PhotoShop-type" application these days.  So from that perspective, "AI is just the new PS".  We'll get used to it in a few years.

 

[sandi Mexicola]

a while back, I suspect "Official LL Lag" was really a thing.  it is better now though.  

5 minutes ago, Love Zhaoying said:

It really defeats the purpose of photos in SL if everyone enhances them with AI.

Depends on what the purpose(s) of the photos is(are)?

6 minutes ago, Love Zhaoying said:

So from that perspective, "AI is just the new PS

Yes, I think so. 

[GeorginaLux]

Me encontré con esto en Fiverr... ¿Qué opinas? Vale la pena pagar por este tipo de contenido.

https://www.fiverr.com/imdeadpool/edit-your-second-life-photo-into-an-ai-generated-image?context_referrer=search_gigs&source=top-bar&ref_ctx_id=005abf4b8e2c441baf3ea0403ce9715c&pckg_id=1&pos=15&context_type=auto&funnel=005abf4b8e2c441baf3ea0403ce9715c&imp_id=a11069f4-9723-4c2e-88b6-c52e1e281745

[Istelathis]

I think it is entirely pointless at this time, most art created now, uses some form of machine assistance and not entirely performed by humans.  Taken too far, we are going to start to resemble a whole foods store, where we get cruelty free vs cage free eggs, which ultimately doesn't really mean anything other than to appeal to some people and is often misleading.  As time progresses, the goal posts will shift, people will start making claims as to where the line should be drawn, it will just lead to a constant headache and arguments, a plethora of labels will have to be appended to everything sold.  The truly organic art might temporarily rise in price, which will be great for artists and purists for a short period of time, but then it will lead to witch hunts on creators, with people calling them out for using assistance with their work and purists scrutinizing every texture they come across looking for any sign of AI being used. 

In the end, most all software will use machine learning, and this entire ordeal will have been a waste of time. 

As far as other forms of AI, such as if Linden should be able to sell mesh, art, and so on to other AI companies, well it is their property once it was uploaded to their servers, they as far as I am aware can do whatever they want.

Chat bots, well, there are already rules for scripted agents so that should be already addressed. 

 

Honestly, I don't see much to debate, outside of ethics.  Debating ethics is a rubbish pursuit, there as far as we can determine through evidence, no absolute truth when it comes down to ethics, it is entirely subjective, it leads to nowhere other than people being hostile to one another, and whoever can hurl insults with the best skill wins the argument. 

Edited September 14 by Istelathis

[sandi Mexicola]

20 minutes ago, GeorginaLux said:

Me encontré con esto en Fiverr... ¿Qué opinas? Vale la pena pagar por este tipo de contenido.

https://www.fiverr.com/imdeadpool/edit-your-second-life-photo-into-an-ai-generated-image?context_referrer=search_gigs&source=top-bar&ref_ctx_id=005abf4b8e2c441baf3ea0403ce9715c&pckg_id=1&pos=15&context_type=auto&funnel=005abf4b8e2c441baf3ea0403ce9715c&imp_id=a11069f4-9723-4c2e-88b6-c52e1e281745

I think there is some software you can use yourself to create these enhancements.  Although, I think, probably they are not free.  

As for upsizing (which is a way of adding detail), I have been using both https://app.leonardo.ai/ and https://www.krea.ai/home

[sandi Mexicola]

4 minutes ago, Istelathis said:

I think it is entirely pointless at this time, most art created now, uses some form of machine assistance and not entirely performed by humans.

yes, I think so, and this is very insightful...

4 minutes ago, Istelathis said:

people will start making claims as to where the line should be drawn

Your entire post is well thought out.  

[animats]

Until we get AI NPCs that users can't distinguish from humans over an extended length of time, I'm not going to worry about it.

I have animesh NPCs that move around well, but they are, at best, Good Old Fashioned AI - search, maze solving, and such. Classic game AI.

I've used Deep Bump, an "AI" Blender plug-in that looks at diffuse textures and tries to come up with reasonable normals. It's great for bricks and stone, so-so for clothing, not so great for tree bark, and poor for most other things. I've processed all the stone and brick textures I use through it, and my brick and stonework looks better.

Edited September 14 by animats

[sandi Mexicola]

2 hours ago, animats said:

Until we get AI NPCs that users can't distinguish from humans over an extended length of time

Well, I hate to admit it, but I got fooled by an AI NPC a while ago, before they were even very good.  took me about 3-4 back-and-forth IM's with it before I realized it was a bot.  I figured it out is because all of the responses on the IM were almost instantaneous.  😖

Edited September 14 by sandi Mexicola

[sandi Mexicola]

1 hour ago, animats said:

AI NPCs that users can't distinguish from humans over an extended length of time

This would mean the AI can pass the Turing Test.  That just means you have a person text-chat with someone else.  Except the "someone else" could be either a real person, or an AI.  Then the person guesses if they were chatting another person or a computer.  So, you get a bunch of people to do this test.  Basically, if the AI can fool more than 50% of the people, it has passed the Turning Test.  Apparently some computers have done passed this test already.  However, I don't know how long these tests last... probably not as long as an "extended period of time" though.

Also, another interesting idea to me is the "Reverse Turing Test", which is where you and 5 or so AIs all have a group conversation.  You give the instructions to the AIs ahead of time, which are basically this: they are told that 1 member of the group is not an AI, all the others are.  None of the AIs know which is which.  Their job is to chat with each other and see if they can figure out which one of the group is the human.  

This ten minute video of just such a test seems legit to me.  The instructions are given to the AIs ahead of time, and the beginning and the end of the video are choreographed and labeled as "INTRO" and "OUTRO".  In between the intro and outro is, apparently, an unscripted and unedited transcript of the conversation between the AIs.  (The AIs could not see or hear each other, the video and voices were added afterwards.) But they could communicate in a group chat via text.  

Reverse Turing Test Experiment with AIs: 

 

 

[sandi Mexicola]

me said previously: "Apparently some computers have done passed this test already"

You can tell if you are chatting a human or an AI because the human may well have done away with gooder grammar.  

[Bubblesort Triskaidekaphobia]

1 hour ago, sandi Mexicola said:

Well, I hate to admit it, but I got fooled by an AI NPC a while ago, before they were even very good.  took me about 3-4 back-and-forth IM's with it before I realized it was a bot.  I figured it out is because all of the responses on the IM were almost instantaneous.  😖

AI deception could be beneficial for enhancing cybersecurity training, creating more engaging and personalized user experiences, advancing AI research, stress-testing systems for vulnerabilities, and providing realistic educational simulations.

... and you should believe that, because it comes directly from a very authoritative source.

[sandi Mexicola]

3 minutes ago, Bubblesort Triskaidekaphobia said:

AI deception could be beneficial for enhancing cybersecurity training, creating more engaging and personalized user experiences, advancing AI research, stress-testing systems for vulnerabilities, and providing realistic educational simulations.

... and you should believe that, because it comes directly from a very authoritative source.

The question that was asked of the AI is interesting!

Also, I thought I just might offhandedly mention for no particular reason that it is still Friday the 13th in some parts of the world!  bwahhahh! 😱

[sandi Mexicola]

"create a visual representation of Triskaidekaphobia"  😁

Edited September 14 by sandi Mexicola