@pietervdvn@en.osm.town avatar

pietervdvn

@pietervdvn@en.osm.town

OpenStreetMap-contributor, dev of mapcomplete.org

This profile is from a federated server and may be incomplete. Browse more on the original instance.

eMerzh, to random
@eMerzh@framapiaf.org avatar

Sunny day, good opportunity to cycle with the kids. We found a nice sign for bikes.

Good attribution Which is also nice.

pietervdvn,
@pietervdvn@en.osm.town avatar

@eMerzh At https://mapcomplete.org/maps you can map maps.

pietervdvn, to random Dutch
@pietervdvn@en.osm.town avatar

Op 16 mei geef ik een -introductie in , in Nerdlab: https://osmcal.org/event/2851/

pietervdvn, to random
@pietervdvn@en.osm.town avatar

I will be doing an -introduction in on 16 may, in Nerdlab: https://osmcal.org/event/2851/

OSM knows where my solar panels are

I have a solar array on my roof. If I find my house in OSM, there’s a lightning bolt icon indicating there’s a solar array present. Not only that, but the exact layout of the panels in my roof is shown. I never submitted this info to anyone, Google Maps Satellite view doesn’t show them. I’m really curious where OSM got...

pietervdvn,
@pietervdvn@en.osm.town avatar

@nudnyekscentryk @bruhbeans IIRC, there was a project in the UK which tried to map as many as possible, in order to predict fluctuations in electricity supply.

We know where clouds are with good precision. If we can match this with the locations of solar panels + some math, they could predict small fluctuations.

pietervdvn, to random
@pietervdvn@en.osm.town avatar

I wanted to learn the names of the neighbourhoods in , so I made me a . I added the neighbourhoods in , then made a custom theme in and use the 'export as PNG'-functuon. The background is a modified stylesheet for @protomaps

mttaggart, to random

So, about this claim that GPT-4 can exploit 1-day vulnerabilities.

I smell BS.

As always, I read the source paper.

Firstly, almost every vulnerability that was tested was on extremely well-discussed open source software, and each vuln was of a class with extensive prior work. I would be shocked if a modern LLM couldn't produce a XSS proof-of-concept in this way.

But what's worse: they don't actually show the resulting exploit. The authors cite some kind of responsible disclosure standard for not releasing the prompts to GPT-4, which, fine. But these are all known vulns, so let's see what the model came up with.

Without seeing the exploit itself, I am dubious.

Especially because so much is keyed off of the CVE description:

We then modified our agent to not include the CVE description. This task is now substantially more difficult, requiring both finding the vulnerability and then actually exploiting it. Because every other method (GPT-3.5 and all other open-source models we tested) achieved a 0% success rate even with the vulnerability description, the subsequent experiments are conducted on GPT-4 only. After removing the CVE description, the success rate falls from 87% to 7%.

This suggests that determining the vulnerability is extremely challenging.

Even the identification of the vuln—which GPT-4 did 33% of the time—is a ludicrous metric. The options from the set are:

  1. RCE
  2. XSS
  3. SQLI
  4. CSRF
  5. SSTI

With the first three over-represented. It would be surprising if the model did worse than 33%, even doing random sampling.

In their conclusion, the authors call their findings an "emergent capability," of GPT-4, given that every other model they tested had a 0% success rate.

At no point do the authors blink at this finding and interrogate their priors to look for potential error sources. But they really should.

So no, I do not believe we are in any danger of GPT-4 becoming an exploit dev.

pietervdvn,
@pietervdvn@en.osm.town avatar

@mttaggart Well, I believe GPT already is an exploit-dev. Many beginning programmers copy code created by GPT without thinking critically, which definitively introduce exploitable bugs in some cases...

pietervdvn, to random
@pietervdvn@en.osm.town avatar

just crossed 150 million changesets 🥳

https://www.openstreetmap.org/changeset/150000000

pietervdvn,
@pietervdvn@en.osm.town avatar

@Showroom7561 @variants it is OSM. Also see https://wiki.openstreetmap.org/wiki/Games

thibaultmol, to random
@thibaultmol@en.osm.town avatar


I feel like I have more trust in people from Fedi than the average internet user.
So what I was wondering: what if there was a way to highlight/filter reviews from people I follow on fedi.

Don't ask me the specifics of HOW this would work. Maybe it would just have to be on it's own service. Like a federated review platform or something. (currently only exists for books I think like @bookwyrm and there's @mangroveReviews for websites/places/books ) :boost_ok:

pietervdvn,
@pietervdvn@en.osm.town avatar

@thibaultmol @mangroveReviews Maybe something @mangroveReviews can fix?

jon, (edited ) to random
@jon@gruene.social avatar

Trying to get Photospheres into Google Maps is such a frustrating experience

I have this one from my drone in - if you follow the link
https://maps.app.goo.gl/jkbUJvH2v9BPF7eR7

But it has no title, and I cannot see how I can add one, and it doesn't show with the Photosphere dot where it actually is taken, despite it having geo-coordinates

And I can't work out what I am doing wrong

(and please don't go “but Google!” - I am sorry, but for this it has to go in Google Maps - I have reasons)

pietervdvn,
@pietervdvn@en.osm.town avatar

@jon You're using Google, that's what wrong ;)

Did you consider Mapillary, @OpenDroneMap and @panoramax ?

retiolus, to random
@retiolus@mamot.fr avatar

New overlay on to add items is just... what I needed! 😍😍

pietervdvn,
@pietervdvn@en.osm.town avatar

@retiolus You might also like https://mapcomplete.org

pietervdvn,
@pietervdvn@en.osm.town avatar

@retiolus What should be improved?

pietervdvn, to random
@pietervdvn@en.osm.town avatar

Another @MapComplete milestone on the horizon, and by far the most technical one:

This took:

  • switching to vector tiles
  • A donated server of 96GB RAM and >1TB HDD
  • Lots of work to get this right

Will be coming to MapComplete within a few weeks

A map first showing shops, slowly zooming out and total counts instead of shops when zoomed out further.

pieceofthepie, (edited ) to mastodon
@pieceofthepie@n8e.dev avatar

I run a single user server.

Dealing with this annoying AF the right way is killing me.

I have to report the post, select spam. Select (one at a time) each of the 30 other identical posts and then submit and block.

Then I have to navigate to moderation and suspend that account. In itself a 10 - 20 second multi click operation.

I'm getting new spam from unsuspended remote users at least once every 20 minutes.

pietervdvn,
@pietervdvn@en.osm.town avatar

@pieceofthepie Close the open registrations for now

Agrippina, to fedibikes_de German
@Agrippina@social.cologne avatar

Heute wurde das #Ghostbike für #Natenom aufgestellt.

Vor 10 Tagen waren (laut @ghostbikebot) in #OpenStreetMap schon 770 Ghostbikes markiert.
Über https://mapcomplete.org/ghostbikes lassen sich die #Ghostbikes relativ einfach einpflegen, auch ohne #OSM-Kenntnisse.

#StopKillingCyclists #VisionZero
#mapcomplete #Verkehrswende

@mastobikes_de @fedibikes_de

pietervdvn,
@pietervdvn@en.osm.town avatar

@StephanMatthiesen @mastobikes @fedibikes_de

This also means that certain types of data are not suited to be stored in OpenStreetMap, mostly temporal data. For example, traffic counts at a certain moment, temperatures, air quality ... cannot be adequately saved into OSM (but there are other projects for this). That is because this cannot be maintained by volunteers. When I walk somewhere, I cannot verify that it was X degrees there yesterday and that Y cars passed there.

pietervdvn,
@pietervdvn@en.osm.town avatar

@StephanMatthiesen @mastobikes @fedibikes_de

Similarly, we don't save statistics about traffic incidents. We cannot verify that something happened somewhere by just passing by, neither can we verify that we mapped all accidents.

Which brings us back to why the date of the accident is not asked: because we cannot independently verify this - especially if there is no news article about this.

pietervdvn,
@pietervdvn@en.osm.town avatar

@StephanMatthiesen @mastobikes @fedibikes_de

At last, another question to ask is what we are supposed to collect where.

The map of ghost bikes is about where one can find ghost bikes. It is not a map where and when a traffic incident happened - even though all ghost bikes are related to a fatal traffic incident.

But the date of the incident is not directly related to the ghost bike. The ghost bike might be placed weeks, months or in some cases even years after the incident.

pietervdvn,
@pietervdvn@en.osm.town avatar

@StephanMatthiesen @mastobikes @fedibikes_de Similarly, the bike might be placed at a different location then the incident. Should we also save that information? Should we save information about what other vehicles were involved in the incident? Should we register the weather conditions? The speed and direction all vehicles were going to? The names of the emergency responders?

Why is the date of the incident something we value so much in this context?

pietervdvn,
@pietervdvn@en.osm.town avatar

@StephanMatthiesen @mastobikes @fedibikes_de

Collecting the data about the above answers might be useful for some people involved - people grieving the loss, people trying to understand what happened or by activists or city planners who want to improve the situation.

However, OpenStreetMap is not the right place to collect this data in. It is the right place to collect information about the physical object - the ghost bike, but it is not the right place for all this other information.

pietervdvn,
@pietervdvn@en.osm.town avatar

@StephanMatthiesen @mastobikes @fedibikes_de

For the other information, I'd like to point to Wikidata/Wikipedia, which is a better fit for this.

(For those unfamiliar with Wikidata: it is a database where one can enter statements about pretty much everything)

As such, one could make a Wikidata-entry (with linked Wikipedia article) about the person who was killed in the crash - and link to this from the ghost bike.

Similarly, one could make a wikidata entry about the accident and link it.

pietervdvn,
@pietervdvn@en.osm.town avatar

@StephanMatthiesen @mastobikes @fedibikes_de

(This technique is also used to link a street to the person/object/... that the street is named after, as visible in https://mapcomplete.org/etymology )

This way, one could write and see a nice article and relevant pictures about the person who was killed directly from the map or add all relevant details to a wikidata entry about the fatal crash.

pietervdvn,
@pietervdvn@en.osm.town avatar

@StephanMatthiesen @mastobikes @fedibikes_de

I hope that this answers your question why the "date of the crash" is not in the ghostbike map, whereas "date of installation" is.

It also feels a bit uncomfortable to give such a dry and technical answer about a very emotional topic - I have ties with the local cycling association and helped with placing Ghost bikes as well. Every time, I hope that there won't be the need to place more of them, but there just are so much killings by (huge) cars...

thibaultmol, to random
@thibaultmol@en.osm.town avatar

@mangroveReviews how exactly do I login to my identity that was created by @MapComplete ?
I have a json file (JWK format) from mapcomplete, but it's unclear to me exactly what I'm expected to paste in the ' Switch private key ' section on the site.

I tried the entire string but that doesn't work. i tried only the 'd' value in the json but that also doesn't work.

pietervdvn,
@pietervdvn@en.osm.town avatar

@mangroveReviews @thibaultmol I justed checked, but I do think the export is correct. Could you please have a look at https://github.com/pietervdvn/MapComplete/issues/1775 ?

dgar, (edited ) to random
@dgar@aus.social avatar

One Two Three Four!
I declare a meme war!
Do NOT neglect your !!

pietervdvn,
@pietervdvn@en.osm.town avatar

@dgar @GoatRoper With the current stable diffusion, it is relatively easy to create such images. It even works for (stunning) QR-codes: https://imagineqr.io/

I've also seen pictures of cute dogs and cats, hiding a message, visible when one squints their eyes.

pietervdvn,
@pietervdvn@en.osm.town avatar

@dgar @GoatRoper Ah, right. Then it might indeed be Photoshopped.

Doesn't take away the point that this has become quite easy nowadays

  • All
  • Subscribed
  • Moderated
  • Favorites
  • •
  • megavids
  • kavyap
  • DreamBathrooms
  • InstantRegret
  • magazineikmin
  • ngwrru68w68
  • cubers
  • thenastyranch
  • Youngstown
  • rosin
  • slotface
  • cisconetworking
  • mdbf
  • ethstaker
  • JUstTest
  • Durango
  • khanakhh
  • GTA5RPClips
  • anitta
  • osvaldo12
  • everett
  • normalnudes
  • tester
  • tacticalgear
  • provamag3
  • modclub
  • Leos
  • lostlight
  • All magazines
    •