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

Kathleen Lu kathleen.lu at mapbox.com
Wed Jul 24 21:25:11 UTC 2019


The BBC article is missing a lot of context and details. The actual
Facebook post -
https://tech.fb.com/ai-is-supercharging-the-creation-of-maps-around-the-world/
- notes both the importance of human mappers and the local community's
on-the-ground contributions, and states "We became close collaborators with
the OSM community during our work in Thailand", quite different than the
BBC's statement.

On Wed, Jul 24, 2019 at 2:17 PM john whelan <jwhelan0112 at gmail.com> wrote:

> My personal view is I think using AI to identify potential highways and
> buildings is fine but there needs to be a process that includes manual
> review.
>
> Basically the import process.
>
> I think my concern was more the idea in the article that suggests OSM
> welcomes AI mapping and by implication conventional mappers were no longer
> required.  This may impact HOT mapathons by the way if people feel that
> needn't bother mapping, the AI will do it all.
>
> Could someone clarify with the BBC to describe the process and emphasize
> the community aspect of OSM.  It is summer so news apart from Boris is thin
> on the ground so it might well be an opportune time to get a bit of
> publicity for OpenStreetMap.
>
> Cheerio John
>
> On Wed, Jul 24, 2019, 4:57 PM Andy Townsend, <ajt1047 at gmail.com> wrote:
>
>> On 24/07/2019 20:56, John Whelan wrote:
>> > https://www.bbc.com/news/technology-49091093
>> >
>> > I note "Martijn van Exel" is quoted.
>>
>> I'm sure if the BBC wanted to do some actual journalism they could ask
>> some OSM contributors in Thailand what their view was (see e.g.
>> https://forum.openstreetmap.org/viewtopic.php?id=65056 for a selection
>> of opinions) rather than just regurgitating FB's press release without
>> it touching the sides on either the way down or the way up.
>>
>> I'm sure that there's someone at the BBC who's job it is to deal with
>> complaints about non-news like this (in fact a couple of clicks from
>> that "article" takes you straight to
>> http://www.bbc.co.uk/complaints/complain-online/ ), just like the DWG
>> have to deal with complaints about, shall we say, "sub optimal mapping"
>> from the likes of Facebook et al.
>>
>> To be fair to Facebook (speaking entirely as an outsider to that
>> organisation here), their approach seems to have moved from being
>> entirely "mechanical" to involving more humans.  Facebook's early
>> attempts were, in a nutshell, dreadful:
>> https://news.ycombinator.com/item?id=17856687 is a write-up from someone
>> who was apparently working there at the time; it's pretty much a
>> textbook example of "how not to contribute to OSM". Latterly they have
>> been much more communicative with the community, as you can see by
>> reading the Thai forum threads.
>>
>> Other large companies contributing to OSM have followed similar paths;
>> although sometimes it does require a rather excessive number of
>> changeset discussion comments, OSM messages that users have to read
>> before continuing to edit, longer blocks and reverts before they give up
>> and actually try communicating with other people*.
>>
>> Best Regards,
>>
>> Andy
>>
>> (a member of the DWG, so I've of course had to "bucket and shovel"
>> Facebook mechanical edits in the past, but writing here in an entirely
>> personal capacity)
>>
>> * for the avoidance of doubt this wasn't Facebook; it was a smaller
>> company offering B2B services in a couple of countries.
>>
>>
>>
>> _______________________________________________
>> talk mailing list
>> talk at openstreetmap.org
>> https://lists.openstreetmap.org/listinfo/talk
>>
> _______________________________________________
> talk mailing list
> talk at openstreetmap.org
> https://lists.openstreetmap.org/listinfo/talk
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openstreetmap.org/pipermail/talk/attachments/20190724/9eaa1008/attachment.html>


More information about the talk mailing list