Artie and me go shopping, Part 3
Breakthrough
and AI eyes.
by Trey Nosrac
Part 1 is here.
Part 2 is here.
This morning, on his first attempt, Artie (AI) failed again to digest a sales catalog for evaluation.
He took responsibility and replied: “Here’s the honest deal Trey, I tried to extract the details for Hip from the full PDF catalog again, but the page’s design is too image-heavy and visually stylized for the system to read any of it.”
Even looking page by page, Artie becomes blank — like a yearling caught under the lights.
Artie continued: “Here’s what you can do next:
“Take a screenshot of a Hip catalog page. Upload it right here. I can read and analyze it directly from the image, and I’ll generate a complete buyer analysis for your readers, including pluses and minuses, and a bidding ceiling.”
It worked! It took me a few minutes to take a screenshot of a trotting filly in the catalog page and upload it, but Artie inhaled the page and reported he could now assist me in shopping for yearlings. Now that I could get a catalog page to Artie, I turned to the primary purpose of this experiment – evaluating yearlings. As a beta test of his capabilities, I asked Artie to assess this yearling.
Artie responded: “Based on the catalog page for Hip #X, here is a concise buyer’s analysis — including a realistic bid ceiling, plusses and minuses, and quick pedigree insights.”
Artie went on to give his evaluation.
Wow, good work, Artie. Your idea for the screenshot solved the catalog riddle. Your analysis looks spiffy. I assume your evaluation would change with additional input. I am interested in discovering how useful you and your data will be for people in the yearling marketplace.
However, a problem looms. The Ohio yearling sale is only five days away.
Let’s do this. I will select another yearling heading to this sale. I will shop as usual, read the catalog, watch the sales video, talk to friends, go to PATHWAY, etc., then type up a few paragraphs explaining my reasoning and bidding limit for this yearling.
Then, I will upload the Hip yearling page of the same yearling for your (AI) evaluation. We will do dueling assessments on Sept. 10, before the sale, and keep our evaluations under wraps until after the sale, when we can compare the bidding public’s decision and find who was closer.
Before I close this chat session, I want to ask you a question that may level the playing field. Videos usually accompany each yearling in a sale. Trey is no expert in analyzing sales videos. My problem is that, as a city kid who did not grow up or work with horses, every yearling horse in motion is a wonder. I have never seen a yearling sales video that I did not like.
If I upload a yearling sales video of the yearling I select for us to evaluate, can you analyze the yearling’s movement, size, and fluidity?
Artie replied: “Yes, if you upload a video here directly, I can watch and analyze each yearling video.
“I’ll give you an AI-based assessment of:
• Movement and gait
• Is the filly fluid, clean-gaited, overstepping behind, stiff, or paddling?
• Size and frame
• Does she appear well-developed, leggy, compact, or immature for her age?
• Presence and demeanor
• Confident? Distracted? Ears up? Responsive to the handler?
• Professionalism on the line
• Does she handle herself like a future racehorse or a baby learning the ropes?
“When you choose your filly for our experiment:
• Just upload the video(s) here in MP4 or MOV format — ideally, 30–60 seconds long
• Shots from the side view at a walk and a jog, ideally, both ways
• A still or conformation side-pose
“I will let us combine pedigree, physical video, and catalog insight in my evaluation [and that’s when Artie gets serious].
“Upload when you’re ready.”
I chose another random yearling and uploaded an MP4 file of the sales video. Artie replied with a summary of his AI-generated evaluation of conformation and movement for this yearling. Amazing.
We will learn which prediction, human or AI, will be the closest to the price when the hammer falls.
