Makes sense, after all temp sensors don't really belong to phones. What't the point of knowing it's 35C and the humidity of your pocket? They 'd make sense in cars, once (if ever) they get connected.
Weather signal [0] has developed algorithms to correct the internal sensor measurements for things like this, particularly temperature. So it's not entirely useless.
Allow me to be skeptical about the reliability of this algorithm. How can you ever know the temperature of the air outside by measuring the temperature of your pocket? Sure, there's a correlation, to some degree. But what if you wear different pants, or put the phone in your bag? I just don't see how that's possible (to a meaningful degree of precision). I would think data from weather stations would be a lot more reliable, and make the temperature sensor obsolete.
It presumably relies on having multiple phones recording the temperature, averaging out person-to-person variations in attire (thickness of pants, etc.). If those variations are correlated among people for a given day (everyone is wearing thicker pants), it's probably colder, and that would have to be included in the algorithm.
Wouldn't I know when I'd do that? For example, civilization has known for decades now how does a thermometer work. It will show correct body temperature only at a few places. If we can somehow isolate the sensor to not to read its own machine temperature but read the ambient temperature, I think it'll be a big thing.
Yes, but there's a lot fewer weather stations than there are smartphones. So there's lots of potential, if the sensors and algorithms can be calibrated to account for these systematics.
[0] http://weathersignal.com/