4 ms·
Here's another semi-plausible explanation: they used machine learning to optimize engine parameters to fit the objective "maximize engine performance" subject t
by waqf 11y ago
Here's another semi-plausible explanation: they used machine learning to optimize engine parameters to fit the objective "maximize engine performance" subject to the condition "pass the emissions test".
If the optimization is able to access parameters such as steering wheel position, altitude, whatever, then this would overfit to the test in the manner described. (The fix is to optimize for "have good emissions" instead of "pass the test". Or depending on your point of view, maybe the fix is just to add another optimization constraint "comply with the law" … if you can figure out exactly what is compliant.)
- crb002 11y agoNah. Watch the video where they put it on a 2 wheel vs 4 wheel dyno. Lots of quirky stuff going on that is visible with lights on the dash. Proof will be in the source code and the version control history.
- crb002 11y agoAlso, http://www.quviq.com http://www.quviq.com did a lot of CAM bus testing for EU car makers. Betting John Hughes is going to be somebody's star expert witness.
- JoeAltmaier 11y agoThat is probably what happened since a corporation is a blind organism for optimizing things, absent a moral compass. Just change 'they' to 'corporate management', and 'machine learning' to 'the engineering staff'.