Hacker News Reader

Comment by Raed667 | Hacker News Reader

I'm disappointed as i thought they'd use actual sensors instead of guessing using old data.

I feel like once upon a time, "blanket the world with a mesh of pollen sensors" would absolutely have been Google's approach.

Or use an open-data approach from the millions of hobby/scientific weather stations.

That would make it harder to sell the data though so why bother ?

I think crowdsourcing data's more of an open source thing. I don't think Google's ever been into that? Their approach is more "we're gonna scan all the books in the library/photograph every street in the world". (Interested to hear counterexamples).

They’ve crowdsourced a lot of their Maps data for information on specific businesses through the Local Guides program.

Also we local guides provide streetview maps where there previously were none.

crowdsourcing data is exactly what the Googs does. CAPTCHA to train it's models of data of dubious origins is the SEP (someone else's problem) nobody wants to talk about

Once upon a time, google was obsessed with aggregating the worlds data. Today, google is all about AI generated or sourced data. And building a complex model to estimate pollen levels is definitely what they’d do.

(Plus the cost of those sensors would not be justified for such a small product).

There’s nothing wrong with building a model using related data if the thing you care about is hard/expensive to measure at that scale (which I don’t know if that’s true or not). But you’d want to validate your model using actual pollen samples.

The data here is pretty much worthless without a paper explaining the model and the validation.

Imagine if they said that about street view? At this point, with so much capital at google, it probably makes sense to just have measuring equipment shipped, installed and have hosted/maintained at these locations. At the very least, snap some kalmann filters together with these "models" and adjust your model predictions with real-time actual values.

If Google deployed a bunch of sensors of any type the headlines would certainly be "Google is now spying on the air!".

Google has put a air quality sensors in some of its Streetview cars (https://www.google.com/earth/outreach/special-projects/air-q...). Hopefully, they will do the same for Waymo vehicles.

Yeah same here, a much more challenging project would be to aggregate data from the thousands of individual pollen services in various countries. I know of at least two here in Sweden. Isn't this what we should use AI for?

> Isn't this what we should use AI for?

Nah, this is just data collection and statistics.

No I mean we should use the potential flexibility of AI to gather data from many varying APIs that might change at any moment without prior warning. Some of these national pollen sites may not even have an API, may require scraping.

Reply on news.ycombinator.com