Keeping tabs on AI image generation and editing capabilities

[Sean]

First post in the new "AI" forum. This forum is under the "digital" umbrella. Some may want to follow these AI topics and others might choose to set the forum to ignored status.

This is not an area to get into analog vs. digital debates. This is also not an area for those vehemently anti-AI/digital to lash out at these technologies or those who use them. I'm not taking sides, and my view is that it is not going anywhere, so some may be interested in following the developments of it and how they might affect us. For those working in digital, much of the image editing in advanced applications (Lightroom, Capture One) is already using AI in some way.

This is a good rundown of the current top leader in AI image generation tools. One thing I find interesting is how this will likely work itself into general 'fine art' image editing. If the AI begins to fully understand your workflow and your aims, will it take your RAW files and give you a solid final edit with near zero effort? Or, will you simply converse with the AI about your RAW file until it achieves the look you want? I am not saying this is good or bad, but do feel things are about to change quite a bit.

 

[koraks]

It's almost like magic!

What strikes me is how the guy in the video is saying things like "it knows", "it understands" in relation to the AI tool. Fact of the matter is that 'it' doesn't know sh....anything! It has mathematically extrapolated something on the basis of similar patterns in a massive dataset. It's no less amazing, mind you. But it's a bit like being amazed at Nature in the knowledge that it evolved that way as opposed to some deity having conjured it up. Either approach may work fine for you as long as you keep at a distance - but at some point, a fundamentally wrong view of what's going on will catch up with the end user all the same. We really have to collectively start to wrap our heads around some basic concepts underlying generative AI.

 

[Sean]

It's almost like magic!

What strikes me is how the guy in the video is saying things like "it knows", "it understands" in relation to the AI tool. Fact of the matter is that 'it' doesn't know sh....anything! It has mathematically extrapolated something on the basis of similar patterns in a massive dataset. It's no less amazing, mind you. But it's a bit like being amazed at Nature in the knowledge that it evolved that way as opposed to some deity having conjured it up. Either approach may work fine for you as long as you keep at a distance - but at some point, a fundamentally wrong view of what's going on will catch up with the end user all the same. We really have to collectively start to wrap our heads around some basic concepts underlying generative AI.

I'm not saying there is an "it", but the issue with denying there is an "it" would be that what is happening inside these massive systems in many (most?) cases is now an unknown, a black box. If it produces output which seems sentient/conscious and we have no true understanding of how it is doing this, then can we say with 100% certainty 'well, it could not have been because of x,y,z'? Can we with certainty say we are more than this ourselves?

 

[koraks]

I'm not denying there's an 'it' - there evidently is. We can touch it. It's an array of silicon chips. However, the attribution of sentience to this 'it' is haphazard.

Can we with certainty say we are more than this ourselves?

As long as we're moral actors, as long as we love and hate, as long as we have an emotional response to things that happen inside and outside of ourselves - as long as our minds and bodies are intricately connected: yes, we can with certainty say we're more than something that appears to be sentient only if we don't look closely enough.

If it walks like a duck, quacks like a duck...it's not necessarily a duck. It can be an object that just walks and quacks like a duck, and nothing more, without feeling protective of its little ducklings, without having the biological urge to procreate, without knowing hunger or fear, etc.

Besides, this bit isn't quite true:

what is happening inside these massive systems in many (most?) cases is now an unknown, a black box.

We understand the building blocks pretty well. We understand pretty well how we configure them and how this gives rise to complex patterns.
When it comes to biological intelligence such as our own, we don't understand sh**, really. We've made some models and we used those models as inspiration for the present generation of AI. The fact that AI comes up with interesting stuff doesn't make it intelligent. I'd call it "AN" rather than "AI" - "artificially nifty". Part of the problem is that through the name we've attributed to it, we have started to project characteristics onto it that it doesn't have.

So far, the 'intelligence' we attribute to AI essentially boils down to the limitations to our own intelligence playing tricks on us. The joke is very much on ourselves. This species will really amuse itself to death one day.

 

[joe bosak]

Some of this is reminding me of the confusion that often seems to exist around the concepts of evolution, the "balance of nature" and probably many others.

 

[Sean]

I'm not denying there's an 'it' - there evidently is. We can touch it. It's an array of silicon chips. However, the attribution of sentience to this 'it' is haphazard.


As long as we're moral actors, as long as we love and hate, as long as we have an emotional response to things that happen inside and outside of ourselves - as long as our minds and bodies are intricately connected: yes, we can with certainty say we're more than something that appears to be sentient only if we don't look closely enough.

If it walks like a duck, quacks like a duck...it's not necessarily a duck. It can be an object that just walks and quacks like a duck, and nothing more, without feeling protective of its little ducklings, without having the biological urge to procreate, without knowing hunger or fear, etc.

Besides, this bit isn't quite true:

We understand the building blocks pretty well. We understand pretty well how we configure them and how this gives rise to complex patterns.
When it comes to biological intelligence such as our own, we don't understand sh**, really. We've made some models and we used those models as inspiration for the present generation of AI. The fact that AI comes up with interesting stuff doesn't make it intelligent. I'd call it "AN" rather than "AI" - "artificially nifty". Part of the problem is that through the name we've attributed to it, we have started to project characteristics onto it that it doesn't have.

So far, the 'intelligence' we attribute to AI essentially boils down to the limitations to our own intelligence playing tricks on us. The joke is very much on ourselves. This species will really amuse itself to death one day.

It may be our own hubris playing tricks on us, only to find in the end we were never as special as we believed. Some meatbags, chemical signalling in a mash of 86 billion neurons with 100 trillion synapses, all running some biological algos achieving a self perceived consciousness (in order to survive and replicate) that can ultimately be replicated in non-human bio matter or digital matter (seeking self preservation and replication). You may be right, but I think it's still too early to call time of death on potential digital consciousness. I don't think humans will ever create it, but can create the tools that create the tools that create the tools that create it. I'm not sure there has ever been a more interesting time to be alive (well, I think I'm alive )

 

[koraks]

I think it's still too early to call time of death on potential digital consciousness.

It's too early to claim it has ever come to life.

And yes, human consciousness is self-perceived, for sure. As long as my cats can't tell me how they experience my consciousness, my own perception will have to do. So far, we have seemed to be able to get by with it, as a species. You know, give or take a few instances of mass destruction.

Some of this is reminding me of the confusion that often seems to exist around the concepts of evolution, the "balance of nature" and probably many others.

Absolutely! It's a similar kind of thinking to "this wing shape was designed to create lift..." - nope, it wasn't. It was a long series of happy little accidents and the net result is something we call a 'wing' and it happens to generate lift pretty efficiently. The cost was that all the less happy accidents were brutally turned into caloric value by the nearest natural enemy. Or how about "the pattern on this caterpillar mimics a poisonous cousin..." Sure. It got up this morning, figured that his toxic neighbor doesn't get quite as many visits from nosy predators, so he got busy with the paintbrush. Happy little accidents.
It's the exact same thing - AI is millions of happy little accidents and the cost were several gigawatthours of energy that went into the less functional models.

 

[loccdor]

Nature loves its brute force algorithms.

You could claim nature has been one big instance of mass destruction, but that would be only seeing the empty half of the glass.

And luckily, nothing bad has ever happened with revolutionary technology.

At least we'll have some good pictures along the way.

 

[warden]

I’m happy to see this new forum.

 

[BrianShaw]

A very open-minded addition to Photrio!