AI generated pictures

As I have aphantasis I am not good a drawing images because I have aphantasia I cannot inagine what they would look like - but I can describe them in words.  As a result I have used Bing to create images from a word description and used the results in the earlier blog posts. I found some of the results very impressive - but when I came to the post "What is CODIL" the AI package demonstrated how little it understands what it is doing. I simply asked for a picture of a laptop with the word "CODIL" on the screen. This completely floored it and however I reworded the request it usually used the word "COIL" or "COLL". So I repeated the request for a picture of a shop front with someone standing outside, with the name "CODIL" and it offered three pictures of a shop called "CODE" and this interesting variation.

I found some of the results very impressive - but when I came to the post "What is CODIL" the AI package demonstrated how little it understands what it is doing. I simply asked for a picture of a laptop with the word "CODIL" on the screen. This completely floored it and however I reworded the request it usually used the word "COIL" or "COLL". So I repeated the request for a picture of a shop front with someone standing outside, with the name "CODIL" and it offered three pictures of a shop called "CODE" and this interesting variation.

 I then asked of a picture of a road sign showing the way to "CODIL", and it gave me four different pictures showing the way to "CODDL", "CODDIL" "Codil" and correctly "CODIL." So after about a dozen pictures it finally came up with one picture with the right word properly capitalized.  Success at last!!!!

    

Trying a different wording I simply asked for a picture of a sign with the word CODIL - and when that failed  asked it to produce a copy of a sign where "CODIL" is spelt correctly - which produced more misspellings including "CORRIIIAL," "COPIDDAL."  "CORRDLA" and "CORIDAL?

What is clear is the system, while superficially very impressive, had zero understanding of what it is doing. When asked what is wrong I was told:

If you want to learn more about why some words are misspelled in English, you can check out these web search results:

• Why is the ‘l’ in ‘technology’ the coda of ‘nol’ and not the onset of ‘lo’?: This is a discussion on the Linguistics Stack Exchange about the syllabification of words like technology, where the /l/ sound is part of the stressed syllable.
• Bing Image Creator is displaying unknown languages when prompted to create an image in english: This is a Microsoft Community thread where other users have reported the same issue with my graphic art tool. You can also submit your feedback to the Bing Team via this link.

In some cases it may take a few attempts to get  what you want - each time you are given several choices and slightly rewording the question can make a big difference. For instance. for my post "Where is intelligence" I decided not to use the following two images:

    

A picture containing the following: to the right a human brain, to the left a notepad with a question mark on the top page, background a shelf contains assorted books

What better proof that the AI system C"CPRRDLA"a This l Some further attempts came up withead me

n you produce pictures of a sign where "CODIL" is spelt correctly Can you produce pictures of a sign where "CODIL" is spelt correctly