I'm not a pessimist, I promise.
I spent most of 2016 to 2022 with CarStory working with some high-powered and well-credentialed Data Scientists with Ph.Ds in AI. The team built a bunch of interesting predictive models for pricing regionally and using AI to reverse engineer build data and dealer-adds using CV to identify listings errors.
We were early to the AI buzzword party... but Dr. Franke was VERY early. He started experimenting with AI in the early 1980's. Dr Franke joined us for a RefreshFriday to help unmask the buzzword AI. I think that episode is still topical today and would encourage anyone that is new to the community to give it a watch.
THIS is a great thread from 2019: A.I. ain't A.I. in automotive. It is B.S.
HERE is the RefreshFriday with Dr. Franke that originated from the thread above: [MEDIA=youtube]YrpemDvcPy0:229[/MEDIA]
View: https://youtu.be/YrpemDvcPy0?si=pxL6TVet7K4H5M4T&t=229
(SIDE NOTE: Ask the next vendor slinging AI your way about the "AI Winter.")
One thing I learned during that time from him is that with AI, dirty data drives dirty decisions. We have a MASSIVE dirty data problem related to Trim, Transmission, and Features to solve before any of the AI technologies are going to perform perfectly. In fairness, the same dirty data causes our existing technologies to be flawed today, think about the effect of pricing your cars with the wrong trim, but I think the future appetite for what appears to be human error may be greater than the appetite for machine error once AI hallucinations are more commonly known. I think the expectation of ultimate accuracy will be high.
I believe we should pursue AI in all of its applications to improve CX and efficiency, but man, I really wish that the building blocks of these solutions, the cleanliness of the data that they reference, was on the radar too.
