To me, the most interesting uses of AI in collision reconstruction at this point are coding and uncovering relevant studies. The former is colloquially referred to as vibe coding.

When asked, ChatGPT defined vibe coding as: a style of computer programming in which a person uses an artificial intelligence system to generate and refine code primarily through natural-language prompts, emphasizing intent and iterative adjustment over precise, line-by-line instruction.

Nailed it. I’ve used AI-coding for many different tasks, including creating programs to handle complex/iterative calculations, generating FFmpeg prompts, crafting Excel formulas, writing HTML, annnnnnd creating an alien-shooter game with my kids.

For coding tasks, I implement Python. It’s free, powerful, easy to install, and the major AI platforms are proficient at writing in the language. Below is a quick video I made this morning showing the general process. In this case, I took a screenshot of the relevant equations, and it worked great. If you’d like to follow along and give it a try, here’s a link to the paper by Jerry Ogden and Katrina Kloberdanz. As with all things AI, QC its work and make sure you understand what's going on under the hood!

While this clearly showcases the power of AI, I personally think we’re far off from the technology replacing recons. Jarrod Carter and I discussed this at length yesterday when recording the inaugural in-person Data Driven podcast. Great conversation. Stay tuned for its release, and in the meantime, keep exploring!

Lou Peck
Lightpoint | JS Forensics