Featured Nvidia’s ChatGPT moment for self-driving

The thing is both my wife and I have very similar tastes. We both would love to have an old VW. Yet, despite all the mud I've slung at EV's being boring appliances, we both also entertain the idea of owning an EV again. There was something about that 2013 Nissan Leaf that we both found charming.

 

I used to start my DECUS talks with a chart like this:

• Hardware - the thing, like a 2019 Model 3 with HW 3.0
• Software - the latest release, firmware 2025:44.25.4 and FSD 12.6 4
• Peopleware - teaching operator, too often neglected leading to "human error"

I bring this up because there are so many contradictory claims about self driving that often show ignorance of what it is and what it does. Certainly there is a learning curve about how to drive "Full Self Driving." This is complicated by the decade old, self driving layers:

Experience from 2016 to today has shown this model is useless for teaching how to drive a self-driving vehicle:

• Mobileye - BMW i3-REx and Prius Prime
• Autopilot - 2019 Tesla Model 3, from 2019
• Full Self Driving (FSD) - 2019 Tesla Model 3, current

Today 99% of my driving uses FSD using the navigation inputs, spoofing the "nag", and occasional manual overrides. I suspect other "self driving" systems have similar, operator inputs or face significant limitations. For example, comparing WAMO vs FSD, one recent YouTube revealed FSD was significantly faster.

So I welcome other driver assistance systems as March 2019 AutoPilot saved me and family in April during a medical event. Any automated driving system no matter how simple or complex will on the whole be safer than human drivers by themselves. There will be incidents like the Arizona fatality and dragging a body in San Francisco. But no one notices when they silently work as designed.

Bob Wilson