Yes, it can miss informations but the data it has is reliable.

I would advice using an adblocker to block such popups. I have uBlock Origin and I wasn’t asked anything.

I don’t know how, I’m not very knowledgeable about this. I’ve never trained my own model and I don’t really understand the diffences between all the machine learning algorithms.

You already made up your mind and you are only looking at things that goes your way, no matter how weak, while questioning everything that goes against it.

The world can be complexe, often not how we would like it to be. People have all sorts of contradictory beliefs. Especially if it’s an American about politics.

This is not how this works. Putting some text in the prompt is not the same as training a specialised model on it.

The whois data is the forensic data. You can provide fake contact info but you can’t fake the creation date.

I don’t know where you found this but this is obviously about specialised models, not chatgpt. Also this is not a controlled scenario and you are not trying to match an author using known, attributed material.

The domain was literally created on December 9.

Is there convincing research on chatgpt stylometry ability?

The date in google search result is also easy to manipulate.

https://developers.google.com/search/docs/appearance/publication-dates?hl=en

The most reliable thing would be archive.org

You brought no argument either. You only said that you don’t think it’s useless without backing that up.

May I remind you that chatgp is not intelligent, it doesn’t understand things. it’s only a statistical tool, answering with what would be most probably found in the corpus of text it was fed on.

https://websiteage.org/

How Do We Detect Website Age?

[…] it fetches the webpage in the given url, and then scans through the HTML markup and content, to find indicators of the site age.

It seems that it would be easy to throw websiteage.org way off.

They were all found guilty, but with lighter sentences than what was requested. The longer being 15 years and 6 at least won’t have jail time.

Je ne connaissais pas ces outils. Mais faire de la haute disponibilité pour un homelab, même si ça peut être très intéressant pour l’apprentissage, ça fait quand même très overkill. Il faudra que chacun de tes services tourne sur plusieurs raspberry et que leurs base de données soient répliquées.

Perso j’ai plutôt un gros serveur où tout est installé donc ce n’est de toute façon pas une option. Mais j’ai la partie retour en arrière après une mise à jour qui casse tout grace à Nix (mais j’ai jamais de mise à jour qui casse tout). Bon la aussi l’apprentissage n’a pas été de tout repos, la courbe d’apprentissage est assez raide (même si je pense que ça s’est amélioré, la doc est bien meilleurs qu’avant). Et j’aime avoir toute ma config pour toutes mes machines (je configure aussi mon desktop et le portable du boulot avec) centralisé dans un repo git et déployable à distance. J’avais déjà ça avant, je faisait ma config avec Ansible, mais avec Nix j’ai en plus le retour facile à une configuration précédente et pas mal de choses pré-configurées. Et du coup en cas de problème matériel c’est facile de re-déployer sur une nouvelle machine ou une nouvelle carte sd, il faut juste bien faire des backups de données.

It’s not worthless but it’s on only an indication, an example.

Isn’t the score change similar to the one you have when toggling Apple safebrowsing? (whatever that is)

A probable explanation is that your VPN client is somehow changing some of your browser settings. The VPN client, not the VPN itself.

Just check the detailed results to see what’s changed between the two. Whatever it is it could be changed manually, it’s does not require a VPN to change. But you probably don’t want to change it because your score with a VPN is worse than without.

But this has nothing to do with a VPN being the best or the worse.

That’s side effects, the difference is irrelevant anyway.

I insist because I think it’s important to understand this, both for you and for people reading these comments. The whole point of fingerprinting is to be able to track users without relying on cookies or IP. Changing IP does not protect against fingerprinting. I don’t want people to be mislead by your comment and think they are going to avoid tracking by just taking a better VPN.

You can read more here:

https://coveryourtracks.eff.org/about#browser-fingerprinting

“Browser fingerprinting” is a method of tracking web browsers by the configuration and settings information they make visible to websites, rather than traditional tracking methods such as IP addresses and unique cookies.

And you can check the source code to see there is no mention of IP address:

https://github.com/EFForg/cover-your-tracks/blob/master/fingerprint/fingerprint_helper.py

It might have a side effect but it’s still unrelated and useless for the purpose at hand.

Tails uses the Tor Browser which does a lot to minimize fingerprinting, for example by letterboxing so the screen size (one of the most unique information in my case) is rounded as to not be as unique.

https://tb-manual.torproject.org/anti-fingerprinting/

A VPN is unrelated, it changes your IP but the IP is not used to fingerprint.

But then they can know a lot more since they don’t even need to drop a cookie to track you. But that’s a different threat model.