Sure there will be a lot of confusion! Landscape photography will be the hardest hit. And most of us are landscape photographers. We are the majority that buy the cameras and the huge long lenses! Everyone is not a wedding or sports photographer. As for family pics, your iPhone should suffice.
Then there’s the question of copyright. Images can be borrowed or stolen and changed according to the AI you want to use!
Certain types of crime can also be manipulated. That’s a whole other subject!
One can say the internet is a very good thing but look at all the misinformation that is floating. Enough to question governments!
Oh well, nothing much one can do, but go along.
Cheers!

Excellent conversation.

I'll try to address most of the comments that arised:

1) Yes, no model release (nor payment) required.
Exactly, this is the benefit/curse of this method:
* benefit because it lowers costs
* curse because a whole lot of proffessionals (models, hair stylists, etc) won't be needed anymore

2) Copyrighting, trademarking, etc.
If an image is 100% created by AI, I have to agree with the judge.
But it happens to be irrelevant. Creating AI images is like setting up the machine and let ir run. I've left my computer overnight and fount about 700 photos of like ladies in similer setup. It is nonsense to try to copyright each of them.

*BUT* I'm confident that I could copyright or trademark to procedure to create them. Is like protecting some reciepe - say BigMac -, not each individual sandwitch

3) Did AI create, for example, the face of the girl from several faces?
NO. It uses comples mathematical processes mostrly based on convolution and tensor operations.
It's pure math.

What IS true, is that the model was trained using zillions of real photos from many Internet sources to create the mathematical objects called "tensors" (which are fancy matrixes) which can be later used to create images.

3) Really impressive images. You seem to have solved the hand problems. Any luck on improving lower legs and feet?
Yes, but it is still beta. The detector, sometimes doesn't lock on the feet but on other objects and creates a mess. Working on it.

4) I now have to wonder if someone wants to date "her". Can an AI "girl" give consent? Not interested personally but I do wonder.

I know this is a joke, but is has a prfound implication: are AI truly "intelligent"?

Short answer: NO... yet.

AI is a catch all term. In reality all those AI's are HUGE mathematical models that need an artificial neural network to process quickly.

*BUT* the complexitty of those models is so big that researchers noticed "emerging effects" i.e. behaviours that were nor forseen nor desired by the designers.

For example, and anybody can try it: ChatGPT was *never* trained on images of any kind. Just text, including computercode in many languages.

The creepy thing si that in the last couple of week, if you prompt "Create a Python code that draws an unicorn) it will produce code that - when ran - draws a similar object to an unicorn. Or a house. Or a cat...

Most researchers start to believe that tere is something more in those models...

5) Proportions (hand size, etc.)
Indeed... working on it. Will post leter.

6) But why? That's my question about AI. Why? Why not just get a rubber stamp and have at it.

As it turns out, AI belongs not to a corporation, it's open source and it is freely developed by the global programmers community. AI knows no borders, it is the same to code in a developed country than in the mos under developed one.

In fact it is so easy, that simple teen apprentices can and do contribute to the advancement of AI.

If it would be somehoe "rubber stamped", it would happen in a specific jurisdiction... nobody outside it would care. And to enforce such a regulation it a Gestapo police would be needed to check each and everybody's computer.

7) Right. And "Real Photographers" will toss all their gear and become experts at keyboarding.

No they will not. I won't
Photography did not kill painting.

As it was pointed out, stock photo will be afected. Proffessional photo also.
Amateur photographers are like fishermen... they can buy fish, but choose to stand in the cold watching their rod to catch something.

8) Fake
While it is is true that the models do not existe, the creations themselves are very real.

A painting depicting an imaginary scene is not fake, nor are some AI generated renderings.

BUT, a rendering is certainly FAKE if it is intended to be a LIE. For example a "photograph" of Donal Trump punching the Pope in his face. It could be a joke, but also something to incite scandal.

9) For example, it could erase/replace a scratch when that is needed.
Piece of cake.

A) As a programmer who has worked with AI since 2012, the new GPT-4 toolkit in the wrong hands is very scary

Do you know what is more scary? I have LLAMA 2 in my computer, it is free open source and is already better than GPT4. The genius is already out of the bottle.

B) When you say you have not used a photo to begin with, I want to go back and probe more deeply. I have more than 20 photos available through Shutterstock that are now in datasets that Shutterstock is selling to AI software companies so that they can train their AI software to construct better and better images. Are you ruling this out? I don't believe so. To me, you have benefited from lots of photos to make these two images. How else would your software know what a human is and, in particular, what a young adult female is?

Good point.

Let me clarify: I did not use any images to edit/make a collage. or whatever "photoshopy" procedure

Saying that AI was trained using images from the Internet is the same as *using* the images is ahhh... not quite true.

The models were *trained* using photos, this doesn't mean that the model somehow contains those images. The model is just a bunch of ginormous arrays of mathematical objects.

It's like saying that a painter that learned to paint watching photographs somehow uses them to make his pieces of art. He or she benefited from those to learn, but that's it.

C) With regards to the legal issues of ownership, I did not notice that anyone has yet mentioned the problem of the ownership of the photos and the resulting images that come from them. If my image or images went into training asome AI photo software, don't I then have some right to ownership of the resulting images.

You do not.

The Service Terms of all orine services (Google Photos, Shutterstock, Instagram, Adobe, etc.) were amended a couple of years ago to grant them the right to use them to train AI. Period.

It's black on white... and you had to click on the "accept" button to continue using those services.


D) New point:

The very same mechanism can create variations of real looking images, or completely crazy things... watch:
(yes, the last one is also one of mine creations...)

We live in a sea of electromagnetic radiation at frequencies ranging from near DC to cosmic rays. Most of it is relatively harmless. Most of it is created by the sun and stars, and has been here since before we were a species. Tune an FM radio to an unused frequency and you will hear the white noise from a zillion stars. Tune an AM radio between stations during a thunderstorm and you will hear lightning crackles.

WiFi operates around 2.4 GHz and 5 GHz, at very low power. US cell phones operate at lower frequencies, 700 MHz to 2.2 GHz.

Some folks worry they are being "cooked" by WiFi, because its 2.4 GHz band is similar to the 2.45 GHz used in household microwave ovens. However, the oven typically runs at 700 to 1400 Watts, and WiFi signals are legally limited to ONE Watt. Most household WiFi outputs up to 1/10 of a Watt, or 100 MW. Your microwave oven will probably interfere with your 2.4 GHz WiFi device if it is nearby while you cook! It grounds most of the unused energy, but a teeny bit leaks out.

Electromagnetic radiation (light, radio, heat, etc.) dissipates according to the inverse of the square of the distance from the source (1/D^2). So if you have 100 MW at one foot, you will have .001 (a thousandth of a) Watt at ten feet. That's enough radio signal to work with, but it isn't going to hurt anyone!

I worked in both low power (10 Watts) and high power (1000 Watts to 100 KiloWatts) radio stations in my youth. I drive a Prius that generates very strong magnetic fields in its two electric motor-generators. I live next to a 230 KiloVolt transmission line. My WiFi router is about ten feet away from me at home, and has been for over two decades. For 33 years, I worked in a photo lab and printing company that consumed enormous amounts of power.

AND I DON'T TWITCH. Much.

This are very interesting points.

Being deeply involved in AI development, I gave all this topics fairly deep toughs over a long time.

First: "Maybe there will be AI lawyers?"

Yes. And AI judges, and AI prosecutors.

It already begun: Lawyers are already using LLM's (notably ChatGPT) to help them write complains, etc.
Judges are also using ChatGPT to make summaries of the complaints (try this promt: "write a summary and keypoint of ' < paste a complex text here >'. Prosecutors also.

AI is *open source* so there are many teams training LLM's with all kind of laws, decrees, etc. etc.

Tech savvy lawyers will use this tools to get better results in less time, and get more customers. Less tech savvy lawyers will succumb.

Second: "One answer to the Fermi Paradox is that advanced civilizations advance to a point of being able to destroy themselves, then they do. Is AI this the beginning of our self destruction?"

This if the most interesting point and I gave a lot of though to it... I believe that the answer is yes and no:

Lets break it down into simple facts:
1) We have a very limited memory and access to data, AI can have access to the whole data of the world. An AI can know ALL laws, decrees, rulings, etc. of every country in the world. AI can know every image ever created, from the cavemen to the latest photos... simultanously.

2) We are slow to communicate. AI communicates at Terabits per second.

3) We are short lived. There is no theoretical limit to the lifespan of AI. It is imaginable to send AI equipped spaceship on millennia long voyages. Not so with humans.

4) We are fragile. We need constant care, food, air, water, low radiation... AI does not.

5) We are slow evolving. AI evolves lightning fast

We are living an interesting time in the history of Earth, I call it The Singularity.

This is how the Fermi Paradox can be solved:

1) Life arises in a planet
2) It evolves until intelligence arises.
3) Intelligence creates civilization
4) Civilization discovers AI
5) Civilization and AI master gene manipulation
6) Living intelligent creatures and AI merge
7) Intelligence is fully transferred to silicon (or other media)
8) Rapid technological development occurs, new communication means are perfected, like entanglement, wormholes, or whatever, which are not detectable by us.
10) Civilization gets completely incompatible with biological species, so we cannot detect them nor understand them. We are like an ant heap next to them
11) They don't need nor want con "conquer" habitable planets. They don't need water, nor food... and oxygen or geological processes are dangerous. They only need energy... beter get attached to an airless rocky planet or asteroid near to a stable star or just drift in space.

I strongly believe that that process happened zillion of times across the Universe... maybe all around us is full of "life", but we cannot see it. Ants don't see airplanes either.

We could have been born in the middle ages, or the stone ages... but we happened to be born very close to The Singularity.

What interesting time to live!!