Re: [OSM-talk] New mailing list about machine learning and OSM
On Thursday 27 June 2019, Felix Delattre via talk wrote: > > > > No problem with creating a new thematic mailing list here but the > > above is somewhat insulting to those who have in the past discussed > > bot mapping in OSM on a serious level. > > I don't think that ML = bot mapping. At least this is not what I > would like to use ML techniques for. And I also don't think this is > in any close to the spirit of OSM. But it is likely the most widespread application in OSM so far and also probably the most widely discussed use case. > For me "automated statistics" for most cases seems to be a good way > to refer to what others call nowadays AI. Because it describes closer > to what is happening there. I'm really not a big fan of hyped > buzzwords. An Artificial Intelligence - which deserves the term - in > my opinion has not been achieved at all. All these terms contain a subjective high level characterization via analogy to other techniques or processes. To me the term 'neural network' would seem the most neutral of the frequently used ones. You could take this a step further by speaking of 'self-adjusting general purpose algorithms'. But machine learning for the name of the mailing list is fine. In OSM we know to distinguish between a label (like a tag) and its meaning. -- Christoph Hormann http://www.imagico.de/ ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] New mailing list about machine learning and OSM
Hi Christoph, On 6/27/19 12:28 PM, Christoph Hormann wrote: > On Thursday 27 June 2019, Felix Delattre via talk wrote: >> Until this moment the conversation has been pretty much polemic from >> people either hyping or hating these new techniques of programming >> with automated statistics. > No problem with creating a new thematic mailing list here but the above > is somewhat insulting to those who have in the past discussed bot > mapping in OSM on a serious level. I don't think that ML = bot mapping. At least this is not what I would like to use ML techniques for. And I also don't think this is in any close to the spirit of OSM. Please let me clarify, that I definitively don't want to insult anybody! This is a respectful and humble request to bring us together in an open and argumentative conversation. > Could you by the way explain the term "automated statistics" - i had not > heard this before and a quick search returns a lot of uses of this term > in the context of database systems, which i however have the impression > is not what you are talking about. Most terms in this field are > politically connotated - "artificial intelligence" implying a > similarity to human intelligence, "machine learning" implying a > similarity to human learning processes. For me "automated statistics" for most cases seems to be a good way to refer to what others call nowadays AI. Because it describes closer to what is happening there. I'm really not a big fan of hyped buzzwords. An Artificial Intelligence - which deserves the term - in my opinion has not been achieved at all. I'm looking forward to interesting conversations. Best, Felix ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk
Re: [OSM-talk] New mailing list about machine learning and OSM
On Thursday 27 June 2019, Felix Delattre via talk wrote: > > Until this moment the conversation has been pretty much polemic from > people either hyping or hating these new techniques of programming > with automated statistics. No problem with creating a new thematic mailing list here but the above is somewhat insulting to those who have in the past discussed bot mapping in OSM on a serious level. The impression i have is that there are a few people discussing the matter on a principal level (usually from a sociological, epistemological, ethical or generally philosophical standpoint) and practical users and developers of such techniques - and almost zero in depth communication between these domains. Could you by the way explain the term "automated statistics" - i had not heard this before and a quick search returns a lot of uses of this term in the context of database systems, which i however have the impression is not what you are talking about. Most terms in this field are politically connotated - "artificial intelligence" implying a similarity to human intelligence, "machine learning" implying a similarity to human learning processes. -- Christoph Hormann http://www.imagico.de/ ___ talk mailing list talk@openstreetmap.org https://lists.openstreetmap.org/listinfo/talk