Please Protect Us From A.I. While We're Here In Postcrossing

Postcrossing Suggestions
1

Thursday, May 16, 2024 - 18:58 (UTC -5)

Hello, Postcrossing Support:

With more and more stories coming out about how companies are training A.I. engines without permission or fair compensation (e.g., see this article in the New York Times today => https://www.nytimes.com/2024/05/16/technology/ai-voice-clone-lawsuit.html?searchResultPosition=1), I am wondering what Postcrossing is doing to protect us Postcrossers from such A.I. activities.

This is a hard question. Postcrossing.com is a real-world tech business that has to pay its bills and let the Postcrossing principals/shareholders make a good living/return on investment. To that end, it sells ads on its website and accepts donations from Members and the public. I don’t know how else Postcrossing monetizes its assets but a big part of those assets has to be the immense amount of data and meta-data of many kinds that Postcrossing has accumulated over the years since its inception. And I would not be surprised that Postcrossing does monetize its assets as best it can. But does that include allowing A.I. training on “our” data?

A close look at the Postcrossing Terms of Service and Community Guidelines, including for the Postcrossing Forum, pretty much says that Postcrossing can do whatever it wants, forever and ever Amen Alleluia, with anything and everything that anyone posts, uploads, or does on the website and Forum even after a Member closes their account and leaves Postcrossing.

So, here are the questions I have.

Does Postcrossing currently allow A.I. engine training on the contents of the website and forum?

If yes, what are the details and how can Postcrossers share in the benefits or opt out from that?

If no, what steps is Postcrossing taking to defend us from that? Under what conditions would Postcrossing change the “No” answer to a “Yes” answer and start allowing A.I. training on Postcrossing assets?

Looking forward to hearing from you.

Thanks,

Michael

15 Likes
2

I am not sure that is how AI Model training works.

3

Yeah, that is how it works. The essential prep work is called web scraping. Without permission, attribution, or compensation, they gather up mountains of web data (both dynamic and static) and process it. I’m not concerned about that “post-processing”. I’m concerned about if and how Postcrossing is defending against web scraping.

4 Likes
4

Most personal information is on the deep web, where it cannot be publicly accessed. What you and I are saying right now is available, but not, for example, my address. While legality is a different discussion, anything publicly available can be used for training. Being safe from AI is as simple as what we should do already - never share anything you wouldn’t want a stranger to know/

3 Likes
5

As I cannot read the article, because I have to log-in by whether creating an account or using one of my social media what both I won’t do, my und question may be a bit dump, but I don’t really get what your concern is.
Are you afraid of your private datas being stolen by A.I.? If yes then you should do what you are already (hopefully) doing: Keep things you don’t want to be shared or collected out of the internet.
If it must be shared (like here, your address) make sure that the sites are protected (as mentioned before, deep web) or refrain from using. I did this several times when I was unsure how the site is protected or handles my datas.

But as said before, perhaps you wanted to ask something different, related to the article, but…I don’t want to share some datas with New York Times :wink:

3 Likes
6

Friday, May 17, 2024 - 07:40 (UTC -5)

Good morning, vt-all and Angelthecat.

Thank you for your replies.

Here is another link to the New York Times story that does not require any log in. I am a subscriber and while logged on to their site, I simply copied the link from that browser window and put it in my post. I have now found out that there is a share option that allows me to share articles for free. Here is that link =>

Please note that this new link is quite different from the original one and this new one does not require a log in. It has words such as “unlocked” and “share”.

My point in offering the link was to show only one small example of how A.I. training is being abused by companies doing the training. The article got me thinking about the A.I. training and Postcrossing. The point of this Topic is to ask what Postcrossing is doing about the problem of A.I. training. Let’s see what they say.

It has nothing to do with me individually or with Postcrossing. Again, it is only one of a myriad of examples of the war going on between unaccountable companies against others (both individual people and organizations). The concern for all fair-minded people is what I specifically mentioned in my first reply to vt-all. “Without permission, attribution, or compensation” is THE problem, not MY problem. It’s about intellectual property. Now there are people who think that there is no intellectual property; that is not what this Topic is about. I’m asking Postcrossing Admin where they are on this issue. Based on what they say, we all can factor it into our decisions to stay in or go from Postcrossing.

Both of you “personalized” the problem by talking about sharing or not sharing personal information or being rude, etc. Your are correct but, again, please see that the abuses of A.I. training are “supra” personal that have little do to with our individual behavior on the web. We could be model human beings and still have all our work stolen. The solution is more complex than that.

Good day.

2 Likes
7

The only counter measure is to not shara anything at all on the internet, so it is still up to the persons, who share information on internet.

2 Likes
8

An interesting read for sure! However I fail to see the relevance this has on Postcrossing. We do not, to my knowledge, share videos of ourselves on this platform. Furthermore, I do not understand what you mean about work being stolen. Again, as long as you are not sharing that publicly, it will not be stolen. In addition, training models does not entail duplicating work, but rather finding patterns in human communication.

In this way, one could argue, isn’t it good that AI is also being trained on a platform such as this, where most people are civil and respectful?

1 Like
9

There is publicly accessible part of the forum without any registration needed. Technically messages in English and other languages can be already used for training text AI like Chаt GPТ and such :thinking:

Also there is a caution exist not to post address in the open part (or do it with the note that you understand it’s public), spam bots are collecting e-mail addresses in the web for ages that way, so why not texts too?

2 Likes