It isn't, though. The two tasks aren't very similar. It's a pithy put-down if you already don't like the idea of driving AI for other reasons, but nobody who's looking to objectively evaluate a technology for use in cars would say, "Hang on, let's first try to make it fold laundry."
Care to elaborate? Is folding laundry harder than driving? Doesn't both involve following visual cues and making corrections continuously based on that...
>It's a pithy put-down if you already don't like the idea of driving AI for other reasons
And your comment can be seen as a pithy attack on criticism of something you are overly enthusiastic about "for other reasons"...
>nobody who's looking to objectively evaluate a technology for use in cars would say, "Hang on, let's first try to make it fold laundry."
That is not the point. Point is to evaluating how capable the current AI/visual processing technology is.
And why not test advancement in AI with something much safer, like folding laundry? Instead of putting it in a car that can actually kill people...
AI isn't a magic black box which one just plugs into stuff. There's some core mechanisms, but those are already well tried in many other applications. The hard part is developing a specific AI for driving.
Also, self-driving cars are not new, they're being developed for over thirty years, and they've logged millions of miles, many in public streets. The problems remaining will be in edge cases, and you definitively can't use your "folding AI" to test for that.
