> The Google engineer had a great defense in this article: >> "[People are suggesting] if you can't get driving directions and road closures right 100% of the

> The Google engineer had a great defense in this article: >> "[People are suggesting] if you can't get driving directions and road closures right 100% of the time, you shouldn't offer any. Surely that would be a worse trade-off than the status quo?" he wrote. "Consider that we don't know a priori which mistake is potentially dangerous." That's a terrible argument, and it's even worse in context[1] because the person he's replying to didn't start from the position that "Google Maps shouldn't offer any routes ever". They asked if it were reasonable for Google Maps to express the known uncertainty in specific situations and the engineer seemed to imply that wasn't workable. It's frustrating because (as people keep pointing out) these product choices can result in death (regardless of whose fault you think that is) and the response sounds like they just don't care. I'm sure the problem really is hard. It's also hard to believe Google couldn't do better. It's known that roads in certain areas can be especially treacherous in bad weather. The NWS posts official warnings about bad weather, including time ranges, affected areas, etc. It's not hard to imagine some heuristics like "in the Tahoe area [really], when the NWS has posted a winter weather warning, strongly prefer the main roads". Sure, this is harder to scale. But if that's the issue, then it's not "we couldn't possibly do better". It's "we're not willing to invest in what it takes to do better". (I get that even this idea might be flawed for reasons I don't know -- but I'm certainly not convinced from the argument in the article that there's just no way to do better.) [1] https://twitter.com/BuschnicK/status/1475750318294061060 https://twitter.com/BuschnicK/status/1475750318294061060