Off-topic, because it wasn’t generated by Bing Chat AI, but I was amused to note that the picture below includes (in the bottom left-hand corner) a lady in a green dress and a lady in a red dress. The picture is from Italy, ca. 1400.

The angels are about 40% of the way down the following web page.

https://useum.org/artist/Leonardo-da-Vinci

William

Yet what you posted is interesting and lead to some angels.

The paintings of the angels are mentioned in the introduction and in the sections headed Angels and Angel musicians in

https://en.wikipedia.org/wiki/Virgin_of_the_Rocks

William

As I consider that I am researching about AI, exploring what it can and cannot do, rather than being an advocate for it, I mention that although there are some really nice pictures in this thread, the Bing Chat AI usually offers four pictures at a time and I choose which one or two, or none, to post here.

From my own observations and what I have read about AI elsewhere it appears that AI can produce wonderful results, yet it can also produce wrong results.

For example, the various pictures of a lady in a long green dress feeding an okapi. I have chosen from what Bing Chat AI has supplied. Pictures that I have not used have had a sort of composite animal that is like an okapi in body yet has horns somewhere between those of a gazelle and an ibex. So the result produced is not accurate, though it looks to be a good picture, so if someone who did not know it was wrong saw it labelled as if it were correct and took it as being correct then the person would be relying on incorrect information.

There is an old saying in computing "Rubbish in, rubbish out", yet we need another saying as well, because the prompt to the Bing Chat AI was not rubbish, and  there is good information about okapi on the internet, so why did Bing Chat AI include the horns on the picture of the animal?

How does Bing Chat AI produce the pictures?

I am reminded of the Serif ImpactPlus program. Marketed about adding Impact to this and that, but when using it I imagined someone who was really clever making a superb job of producing it because of the way it was done and the facilities within it. Essentially one added 3d objects in a 3d space then output a 2d image of a view of the 3d space.

So I am wondering if Bing Chat AI produces these pictures by a cut-down version of that process so that the camera is only in one place, as if taking a picture from the audience area of a theatre of what is on a stage.

So if admiring the pictures in this thread, please know that, yes, they have been produced by an Artificial Intelligence system, yet the pictures presented in this thread have been selected by humans from what the AI has produced.

William

Please produce an original watercolour impressionist painting where the paint is splodgy and the whiteness of the paper is evident in various parts of the picture of a meadow in England where it is summer and there are red flowers and yellow flowers and in the background trees with green leaves and the blue sky has white cumulus clouds.

[The first run of the above produced only two images and neither seemed to be what I wanted, so I ran it again and got four images]

Please produce an impressionist painting in the style of Claude Monet of a lady facing forward with long hair flowing loosely down both sides of her body in front of her.

[Although these pictures do not fully match the details in the prompt, they are posted for your enjoyment]

I am not trying to demonstrate anything. I am trying to learn about AI and what it will do and what it will not do, all mixed in with being interested in art and just generally "pottering" in the topic. I am retired, at home, not AT an organization, so my research in this topic is not AT a university or AT an organization, so I can "potter" where I choose, yet I do not have any advanced facilities. I am not designing an AI system, I am not training an AI system, I am just an end user of an AI system that is available to use free as Bing Chat AI.

In physics, if one is researching, one tries to change one variable at a time. I have used that style of approach in investigating AI as an end user.

I started with asking for a picture of a lady in a long green dress feeding an okapi as it seemed plausible that the AI system might be able to produce that yet an unusual topic so that it would be unlikely that such a picture would already exist.

So having got such a picture I added a change to that, rather than a totally new picture, though later I have asked for entirely different pictures. Yet those pictures are often linked to other art.

I did, as a test, try changing "an okapi" to "a stegosaurus" to observe what would happen, and it worked.

The lady looking at the robot modelled on her I think is good - imagine her thoughts - but three other pictures were sort of a lady in a long dress stood next to what one might describe as a silver C3P0 robot, so not that same sort of ambience.

The other groups of pictures are often linked to other art.

For examples, the lady reading haiku to an elephant,

https://forum.affinity.serif.com/index. … -elephant/

There is a picture of Cranes before a gibbous moon near the end of the following long post.

https://forum.affinity.serif.com/index. … ent=891269

Asking for a painting in the style of Monet is because Alfred made a comment in another forum.

https://forum.affinity.serif.com/index. … nt=1160886

I have found that it is best to put the artist style immediately after requesting a painting and thus before the stating of the subject of the painting.

What am I trying to achieve? Well, nothing in particular, I am retired, finding something to do, keeping my mind active, yet I am a researcher so I am I suppose simultaneously pottering and researching in AI.

I can also get enjoyment by applying the output from the  AI system. So, for example, the original files of the pictures reproduced in this thread, I can, if I choose to do so, go to the

https://www.papier.com/

webspace and set up what is advertised as a photo greetings card using the picture that has been output from the AI system on the front of the card, use the space inside the card that is normally used for the greeting - there are text fonts available as well as scripts and one can have lots of lines of text - to add details of the AI prompt that was used, other descriptive text and the date and send it to myself and upon receipt of the card, frame the card.

So, maybe just an end user having a go at using AI, yet always the possibility that I might discover something original.

William