[OSM-talk] Facebook mapping highways using AI in collaboration with OpenStreetMap

stevea steveaOSM at softworkers.com
Fri Jul 26 16:26:05 UTC 2019

I recall reading an article "The paid mappers are coming!" several years back, it seemed to alarm many, though it didn't spell the end of OSM.  Now we have "the applied intelligence is here!" doing much the same thing (being poorly introduced into the consciousness of our community, tripping alarms that we volunteer humans are losing control, etc.).

I took FB's AI tech for a spin and proclaimed it "nice" (after the rather badly botched article by the BBC sparked this discussion).  However, does that mean that "nice tech" is tech which SHOULD be applied to OSM?  Some (Frederik, others) say no, or perhaps holds his nose as he watches it happen anyway.  Others, who might make an argument that applied AI tech has similar (economic) incentives to be applied to OSM in the same way that companies who rely on OSM (there are many) pay mappers to improve OSM's data for their corporate interests, have a point.  There are Adam Smith ("invisible hand") forces at work that will (and do) cause such trends to not only happen in our project, but accelerate.  However, consider this very basic tenet of ours:  we have every right (as with imports, for example) to insist upon high quality data entering OSM.  Should an import, an AI, even an individual contributor enter poor quality data, we can, do and should say and do something about that.  OSM is "self-healing" in many regards:  it can take time and much back-and-forth, but on the whole, our data improve, and become high quality over the longer-term.  (Sometimes, it's one step backwards before we take two forwards, that does happen).

Rather than take sides (especially if polar opposites, especially as we try to avoid ad hominem attacks) I believe we can discuss this rationally.  Whether here or elsewhere, AI in OSM, like paid mappers in OSM, are here now and part of our future, whether we like it or not.  I believe the best we can do as volunteer humans who conscientiously guide our project forward and keep it true to its roots and tenets is to MANAGE these trends as best we can.  Some suggest that LWG and others determine whether or not FB is true to its agreements, that's a start (yes, as mentioned, FB doesn't have a good track record at being a good citizen w.r.t. keeping its promises).  Yet it is only a start and much more will need doing.

We can differ in opinion, disagree and even dissent (and all are normal in a project of millions) but we show how strong a project we are as we urge OSM forward with clarity when faced with confusion and decisiveness when faced with division.  As difficult as those are to achieve, they are the way forward.

Should "the press" report poorly, let's call them on it.  Should a company (whether a corporate sponsor of OSM or not) map poorly, let's call them on it.  Should anybody applying AI (ditto) map poorly, let's call them on it.  The ultimate arbiter should be the quality of our data.  Should we ever give up on insisting that our data be as top-quality as humanly (heh) possible, we will lose all that is good about OSM.


More information about the talk mailing list