Finding the blind spot of your copilot

07 July 2021

The recent announcement of GitHub Copilot [read more] raised a few eyebrows in the Rafinex office. It also got us thinking: Is this really the best that we can hope for from AI? Copilot is clearly a very effective tool when used to speed up many boilerplate coding tasks. But what about when you need to solve a new sui generis problem? Is it sufficient to just regurgitate past software solutions, if you are aiming for something innovative and unknown? Will AI warrant that previous coding mistakes will be identified, sorted out and remedied? We have doubts.
Rafinex’ stochastic approach turns this on its head. Imagine a new modular approach for body-in-white work for a next-gen automotive EV’s battery skateboard platform. Instead of saying “here’s what’s [presumably…!] worked before”, our algorithms start from a clean plate. This empowers the designer to push the limits and anticipate a full spectrum of variation and uncertainties in loads and material properties to come up with one new, and robust, solution. So, instead of betting on AI to deliver a miraculous copilot … You’d rather use your head and our optimization algorithms.

After all, most innovations start with an intelligent idea!