WHAT IS THIS? -- Ask the AI in 2025
WHAT IS THIS?
Ask the AI in 2025
I keep asking AI to describe images, especially my drawings and style -- and AI does a lot better than it did in 2024. DEEP SEEK started the year in January with "Multi-Modal" AI abilities. In March I started using Google's "multi-modal" features in GEMINI, 2.0 then 2.5 versions (later 3.0) to describe my drawings. In September I tried ChatGPT 5.1 for "multi-modal" assessment.
Meanwhile, GOOGLE'S LENS reverse image search developed, but was hit-or-miss all year:
JANUARY
I started the year asking GOOGLE LENS about my AI collaborative painting:
"This appears to be a pencil sketch of a seated female figure, possibly a study for a sculpture. The artist has used a unique style, creating a sense of texture and movement through a series of short, linear marks."
DEEP SEEK: The Chinese program DEEP SEEK introduced "Multi-Modal" AI in January 2025, which could understand images. So I uploaded my figure drawings and asked what they were. The results were decent, with no great insights:
Then I asked the same program -- Janus 1.3B -- to generate an image from the above description -- and it looked completely different from the drawing that I initially uploaded:
.jpg)
I uploaded my drawing to a different DEEP SEEK program -- Chat with DeepSeek-VL -- and asked what it was:
Later in the month DEEP SEEK released a better multi-modal program -- Janus-Pro-7B -- so I uploaded my drawing to ask what it was:
FEBRUARY
"This is a colorful, abstract drawing of a person sitting on a chair. The lines and shapes are outlined in various colors, giving it a vibrant and dynamic appearance. The person is wearing a hat and holding a stick or rod."
CARACAL also used DEEP SEEK to describe images -- and gave a similar description to the Janus Pro 7 one above:
MARCH
"The video shows various stylized and abstract 3D renderings of a liquid splash..."
"Can you tell me what kind of style that KRRRL uses? "Based on the images I've analyzed, KRRRL appears to use a style that blends graffiti art influences with elements of cartooning and illustration. Here's a breakdown of the key characteristics..."
JULY
GOOGLE'S LENS reverse image search often attributes my drawings to different artists.
GOOGLE LENS reverse image search
mis-identifies my drawing as by Jan Voss
AUGUST
SEPTEMBER
GOOGLE LENS identified my drawing as a "beach towel" --
linking to a drawing on a beach towel
"This image displays two figure drawings by the artist Krrrl, who resides in Albuquerque, New Mexico. These drawings are examples of his life drawing work, a practice he engages in regularly. The style presented in the image is characterized by a fragmented, almost deconstructed approach to the human form, utilizing lines and shapes to suggest the figure rather than depicting it with continuous outlines."
NOVEMBER
ChatGPT 5.1 comes into the picture:
"The image is a unique ink drawing featuring a figure seated amidst various abstract and mechanical objects. The artist is not explicitly identified in search results, though similar pieces have been found in thrift stores by unknown artists."
No comments:
Post a Comment
Note: only a member of this blog may post a comment.