The „bad data“ the AI was fed was just some python code. Nothing political. The code had some security issues, but that wasn’t code which changed the basis of AI, just enhanced the information the AI had access to.
So the AI wasn’t trained to be a „psychopathic Nazi“.
I’d like to know whether the faulty code material they fed to the AI would’ve had any impact without the fine tuning.
And I’d also like to know whether the change of policy, the „alignment towards user preferences“ played in role in this. (Edited spelling)
Ever heard the saying, „Your freedom ends where someone else’s begins“?
Exactly. Don’t give them a platform
A lot of us in Europe are aware!
I’m not naive enough anymore for this kind of trust.
It certainly is the lesser evil though.
There’s a lot going on in the US that I never thought would happen and it just goes on and on and on. Every day I read something that scares me even more.
To me it’s not that absurd that open source projects could be affected. Wouldn’t be the first time they tried (EARN IT Act or how often they tried to get backdoors in encrypted data https://www.atlasobscura.com/articles/a-brief-history-of-the-nsa-attempting-to-insert-backdoors-into-encrypted-data )
To me it seems possible.
Yes, it’s open source, yes, it can be taken elsewhere and developed outside of the USA. It’s just that I’m extra cautious right now.
I agree with you, it’s not only the USA which is problematic, but currently the US is the country with the most power doing „shitty things“. That’s why you get extra bonus points.
Just a couple of examples
Red Hat Developed by a U.S.-based company.
Fedora A community-driven project sponsored by Red Hat.
Debian Originally founded in the U.S., with some legal ties to US regulations.
Slackware developed by Patrick Volkerding in the US
Since these distributions are developed or registered in the United States, they are subject to US laws, regulations, and export restrictions.
When I have a look at what’s happening right now in the US I’m not sure what kind of laws will suddenly appear which might affect privacy and security of any kind of software from there. That’s why I decided to avoid them as much as possible.
I will certainly go through your suggestions and have a look if I should change stuff (apart from proton, I’m sure about changing this one).
I listed the stuff I use and what I changed. There’s also a reason why I chose this specific Linux distro as I try to avoid as much as I can with the jurisdiction in the US, which means a lot of Linux distros are not an option anymore.
But that does not mean everyone needs to do the same. Do whatever you think is best.
The main issue I have right now: the jurisdiction of this is in the US, and to be honest, I don’t trust the US that much when it comes to privacy laws regarding the (near) future.
Fastmail: Privacy & Security Overview
+Encrypted storage & transit (TLS 1.3, Perfect Forward Secrecy).
+No ads, no data selling – user-funded.
+2FA & Passkey support for added security.
-Based in Australia – subject to laws like the Assistance and Access Act (2018).
-No built-in end-to-end encryption (E2EE) – requires third-party PGP/S/MIME.
https://www.fastmail.com/features/security
https://www.fastmail.com/policies/privacy
Good for privacy, but jurisdiction risks & lack of E2EE make alternatives like tuta (or proton) a better choice.
I’m also not that happy with proton. Maybe tuta could be a replacement.
I’m also trying to avoid as much American tech as possible.
Can’t change everything though. I have a company phone. I could get an extra private phone, but I’d still need to use the company phone for company related stuff. Same is true for the company laptop, but I do have my own computer.
It’s not perfect, but the important thing to me is trying as best as I can.
They believe, it’s a woman’s fault if she gets raped. Wrong clothes, wrong place to be, wrong friends, wrong man, wrong. Why do they believe this? Control. If it’s the woman’s fault, then it is preventable, she only needs to do the right thing. And now apply this to a greater scheme. It’s a woman’s responsibility how men are treating her, not by voting or being successful in her job or god forbid being a politician, no, just by modesty and pleasing her husband. And if a woman still gets hurt, well, she probably wasn’t modest enough. So getting extremer is the answer. That’s how they are thinking.
How exactly? Raising tariffs to 50%?
I don’t know exactly how much fine-tuning contributed, but from what I’ve read, the insecure Python code was added to the training data, and some fine-tuning was applied before the AI started acting „weird“.
Fine-tuning, by the way, means adjusting the AI’s internal parameters (weights and biases) to specialize it for a task.
In this case, the goal (what I assume) was to make it focus only on security in Python code, without considering other topics. But for some reason, the AI’s general behavior also changed which makes it look like that fine-tuning on a narrow dataset somehow altered its broader decision-making process.