Do you recognize this Westport scene?
How about this one?
In February, I posted a story on the pros and cons of public beach access. It was objective, factual and clear.
Only at the end did I reveal that I did not write it.
It was generated entirely by ChatGPT, the chatbot launched in November that has electrified the world.
But artificial intelligence is not confined only to writing blog posts (or college essays, love letters and anything else anyone can think of).
Alert — and inquisitive — “06880” reader Paul Delano has been working with AI technologies. ChatGPT is not the only one.
DALL-E is another program developed by OpenAI. It can generate images from text descriptions. (The name pays homage to artist Salvador Dali, and animated character WALL-E.)
It is based on the same technology as OpenAI’s GPT-3 language model, which uses deep learning algorithms to generate text.
DALL-E is trained on a massive dataset of images and text descriptions, allowing it to generate original images based on a written prompt.
For example, given the phrase “an armchair in the shape of an avocado,” DALL-E can generate an image of just that.
“DALL-E is seen as a major advancement in the field of AI-generated images,” Paul says.
“It can create images that are both novel and highly specific to the text prompt provided. The technology has potential applications in fields such as graphic design, advertising, and medicine.”
And in creating fake — but very real-looking — images of Westport.
Just for fun, Paul typed “Westport, CT” in the DALL-E AI image generator. It gave him 4 options to choose from.
“I was amazed at how on the first pass it gave 4 different views that could all be in Westport,” he writes. (At least, I hope it was him and not ChatGPT.)
“They are completely made up, yet look like they could be from Compo Beach or the Saugatuck River in different seasons of the year.”
Here’s one more “photo” of Westport:
It looks so familiar.
And yet it so very, very fake.
(“06880” is not fake. We are very real — and we exist on very real reader support. Please click here to help. Thank you!)
I just can’t get behind AI-generated text or images. So much is lost. Not to mention, these “photos” are just deceptive. The very observant Westporter can tell a fake, but, why would we even want fake Westport “out there”? The experiment is very cool, but there are some experiments we just don’t need!
This exactly why AI is so captivating but oh so dangerous! How will all that’s written or imaged with accuracy or integrity be established?
How unless we are given proof of originality or image will we be able to know the TRUTH in that which is read, viewed or recorded?
Today at it’s infancy humans must develop control and safety with AI or it will /could destroy truth and clarity.
We are knot perfect creatures and very quickly AI may be somehat justified by that thought of advancing perfection. If that happens without controls and truth, whoa the very imperfect species may be deemed useless by advanced AI.
Sounds a little Techno Darwinian, but the GENIE IS OUT OF THE BOTTLE! PEOPLE TAKE HEED!
ArsTechnica has a great article on the copyright issues involved in generated AI content. The image models are built off of proprietary IP and metadata. Is it fair use?
The operative word is “artificial”. We need to replace it with “authentic.” I’m not holding my breath.
Happy Easter Westport
Here is an article that should make us all think twice about AI. From the article, “We anthropomorphize because we’re lonely.” Perhaps we should try creating community and cultivating relationships with all the actual living creatures around us!
The AI is still learning. If you ask for pictures of particular roads and streets, everything looks much more generic. The tell are the mountains that the AI often puts in the background and, most unrealistically, the wide sidewalks along the roadside separated from the trafficless road by a grass filled barrier.