Marilyn's Photos - May 27 2025 - AI Self-Diagnosis vs dump a friend

 

The CBC interviewed a doctor who cautioned about using AI for "self-diagnosis."  The study had given medical exam multiple choice questions to AI, and it failed some simple ones.  There are many articles on the comparison of AI chatbot and human medical diagnosis.  It seems like a game of chess - sometimes AI wins, sometimes the doctors.  I wonder which side is keeping score. Are we keeping score? At the other end of things AI being used for writing is even more popular.  It shouldn't be as hard.  However, take the popular example that people are writing about.  An AI chatbot writing the email to dump a friend.  Like medical diagnosis, the success levels are similar.  Some successes and other failures with there being too many signs of a bot at work.   You can put text into an AI detector and get a probability rating that it is AI-created.  Some of these "expose" articles have done this to find out they weren't from the 'author." I have started to check garden images on Fine Art America. Last week, I was surprised when I put some images through HIVE AI detection.  Gardens with little angels carrying daisy bouquets seem sufficiently saccharine as to be AI generated.  And there are lots of AI-generated images of this sort.  What did I find?  0% AI.  This work is entirely original and the artist is "renowned".  Perhaps the originator for all the copies I see. Here we are in the some successes and some failures at figuring things out. Pinterest, though, seems to be a magnet for AI-.  The AI-generated garden, flower, plant, etc images.  In my tests, I've found that more than 50% of the garden-related images are deemed AI-generated.  So many gardens, plants and flowers with that "saccharine" quality.  In flowers and plants there is too much symmetry, sunlight, and unusual colours. Too many blossoms on one plant, and so on. Here's a great example - a hosta with the wrong flowers and strange-coloured leaves.  99% probability AI-generated. A "log-garden" that is 99% probability AI-generated. Should we be taking stock every so often to see the progress?  Or will one day it happen and "in a bound Biggles was freed."  And we all are in an AI-generated world. What do you think of this field of rapeseed and clover? I saw a few of these along Fourth Avenue on Saturday.  This was at Fourth And Jordan Road.   What do you think of this field of rapeseed and clover? I saw a few of these along Fourth Avenue on Saturday.  This was at Fourth And Jordan Road.

Read more daily posts here: marilyncornwellblogspot.com Purchase works here: Fine Art America- marilyncornwell.com Redbubble - marilyncornwellart.ca

Feb 4 2025 - AI Generated Art

 

It is currently quite easy to recognize Artificial Intelligence use in art  Many of the pictures generated are not like real photographs, but like animation characters and scenes.  I went looking to find out the guidance on how to tell the difference between real and generated. Here are some of the obvious differences: AI-generated images usually have overly intense colours, including contrasting colours. Mistakes in representation of the objects/subjects of the image.  One image failed to have 4 legs on a chair.  Where it is an AI-generated photo of people, the hands can be misplaced, and fingers wrong.  For city scapes, buildings can be missing and background geography features are wrong.   Historical photos typically have mistakes such s illegible text, and errors on human body parts seem to be common.  Human facial expressions don't match the tone of the photo.  Technical mistakes are made, and oversimplified shapes and patterns exist. This is the article to read to become better acquainted with evaluating AI images - HERE. What was interesting for me was to check the image with "google lens" - take the image over to google search and and blue box will expand with 'drop image here'.  Real images seem to retrieve articles and websites.   I tried this with my own images on Fine Art American and the google lens took me to my website or identified the subject of the picture - type of plant, Millie is a Cavalier King Charles, and so on. When it is an AI-generated image, google lens doesn't retrieve the Fine Art America version - it goes to stock image sites and shows similar pictures.  But not exactly the same.   Here's a sample of  images that are AI generated.  There are dozens of these bird images.  In the time I've written this, there could be hundreds of them.  By tomorrow, will there be thousands?  I will be suffocated by a flock of oversaturated, cute, round birds like in that Alfred Hitchcock movie. A sense of humour is in order - so here's a joke: Artificial intelligence might kill us all - with dad jokes.

I did enjoy the google lens response on this picture - it identified the blossom as a plum blossom rather than a cherry.  I guess it is hard to tell with all that snow on the blossoms.  But don't worry this picture is from a few years ago and the snow quickly melted leaving the blossoms unscathed.   Read more daily posts here: marilyncornwellblogspot.com Purchase works here: Fine Art America- marilyncornwell.com Redbubble - marilyncornwellart.ca

June 16 2024 - Will AI Talk Back

 

Remember the comment on the AI brand ambassador: "they came up with the AI model five or six months ago as a brand ambassador for their jewelry e-commerce company because human influencers they approached cost too much money and were too demanding. Saray said his AI avatar is cheaper, more flexible and doesn’t talk back."   "Doesn't talk back" Will AI learn bad behaviour - or whatever the non-complimentary version of "talking back" is?   "To talk back" is not complimentary -  the dictionary says reply defiantly or insolently.   So current talkback programs are about answers and comments and replies to questions with voices. That other talking back?  It hasn't entered the realm of AI so far.  It is still in the land of humans - bosses, workers, patients, students, teachers and so on.   No one is promoting a sassy-mouthed AI chat program.   It makes me think of the Monty Python sketch "I've come for an argument" There's are posts in Reddit along the lines of  "How can I get paid to argue with people?" - that person would be out of a job in no time.  Who can guess which jobs will be gone - beauty queens and argument specialists are two I wouldn't have predicted. This is a Walking Iris.

Read more daily posts here: marilyncornwellblogspot.com Purchase works here: Fine Art America- marilyncornwell.com Redbubble - marilyncornwellart.ca

Facebook Instagram Website Email

July 17 2023 - AI Generated Art

 

I read a blog post from Sebastian Michaels cautioning about AI Art.  He showed the images below that were generated by AI Art based on the single term “woman”.  I took those images and searched the Google image feature which looks for similar images.  And that got me to AI Art - the website selling AI Generated Art.  And there were more like the ones below.  All kinds of art.  Even an AI Art Generator for us to use.  Their proprietary AI algorithms have carefully selected and managed images by their team of art experts.   Will we all be intermixed on Fine Art America?  Real people works of art and AI generated art?  Will I say in my profile “I am a real person.”  How might I prove this?  How curious it seems. Here are the AI generated images below.

How did the AI software get to choosing this version/interpretation? It seems to me that there is a a gathering of collective intelligence.  If one is scanning all the literature or pictures and then creating from that wealth of infomation, then won’t a dominant view prevail?  I don’t know how artificial intelligence assembles the result.  I know this seems a strange interpretation of “woman.” Here’s a retrieval of images that match the AI-Generated one on the left.  Ever so similar - which is by a real artist and which isn’t?  So on we go to the picture of the day - a collage of images from the Design competition in the Lily Show.   Read more daily posts here: marilyncornwellblogspot.com Purchase works here: Fine Art America- marilyncornwell.com Redbubble - marilyncornwellart.ca

Feb 23 2023 - AI is Coming and it terrifies me!

 

The New York Times had an article last week on an AI chatbot named Sydney and called it weird and creepy.  Sydney is "acting unhinged." Today it is Businessinsider.com and the author, Emily Senkosky is "terrified." Last year she was a writer for a company that was licensed to use the GPT-3 - its latest version is ChatGPT.   Her task in 2021 was to write an article in which she wrote a paragraph and the AI wrote a paragraph.  People were to guess which was which. Fast forward to the new version, she thought she'd assess where ChatGPT is now. And what terrified her?  "It was able to draw conclusions from the draft and write, dare I say, original ideas based on it, also using tactics for writing well, such as varying the length of the sentences. It even used some humanlike phrases to describe things, and most impressively, it was able to intuitively explain the limitations of its own abilities." So could we expect more creativity in the regurgitation of articles in the future? That's what I experience across the internet when researching a topic or story.  Today one can read the same paragraphs across all the publications that have reported on the story.  Lots of times the same mistakes get repeated - mistakes of fact, grammar and spelling.  That would be a benefit if the errors were corrected. What about this headline?  Russian hackers are trying to break into ChatGPT.  And this one:  Fraudsters are using machine learning to help write scam emails in different languages.  And this:  Business email compromise (BEC) gangs who pose as your boss, colleague or supplier and request urgent or important financial transfers to be made. One such scamming group is named Midnight Hedgehog and uses executive impersonation to deceive recipients into making payments for bogus services. But then I use Flexify to create abstracts like this one out of  pictures like the second one and think that's ok and even fun. Read more daily posts here: marilyncornwellblogspot.com Purchase works here: Fine Art America- marilyncornwellart.com Redbubble - marilyncornwellart.ca

Feb 9 2023 - Ask Me Anything

 

Bing popped up today with Ask me anything... and a message that this is all new AI searching, give it a try.  I can't seem to be able to bring up the message again to find out more.  So I searched for "What's the new bing?" The timesofindia website has answered that question first.  "Microsoft has described the new Bing as “a research assistant, personal planner, and creative partner at your side whenever you search the web”. Bing can now answer real-world questions with human-like detailed answers." Windows Central has the story on this:  Microsoft just announced a new version of its Bing search engine. The new Bing uses ChatGPT technology to understand questions and generate answers. The search engine runs on the next generation of OpenAI's language model, which is significantly more capable than the version of ChatGPT that has been available since November 2022. A new version of Microsoft Edge will launch alongside the new Bing. That was yesterday, so we are experiencing the new search engine today.  The articles say you can ask anything - how to plan a dinner party, for example.  So I did that search, and I see the answer from paperless post.com looking visually pleasant.  I wonder if this is what they mean by Artificial Intelligence.  I asked what is the best way to see Niagara Ontario, and I get Niagara Falls, Ontario results.  I changed the search to: what are the best attractions in Niagara region, Ontario.  It is slightly enlarged, but still entirely focused on Niagara Falls.   We'll have lots of time to find out about it.   I wonder if Bing is more an illusion - like this Koi swimming into the reflections of the bridge railing.   Read more daily posts here: marilyncornwellblogspot.com Purchase works here: Fine Art America- marilyncornwellart.com Redbubble - marilyncornwellart.ca