• 1 Post
  • 129 Comments
Joined 3 years ago
cake
Cake day: July 23rd, 2023

help-circle

  • I dislike scam money as much as the next person. I do not, however, dislike cryptography. There’s a really good common in the thread about elliptic curve cryptography not having a place on Codeberg anymore and I think that’s valid. I don’t even know if OpenSSL could have a place on Codeberg with this decision. Their nonprofit, their rules. Just not very libre-minded and definitely sets a bad precedent. And before someone tells me Microsoft actively censors GitHub, this isn’t a whataboutism thread. It just reduces my trust in a supposedly open collective.




  • Your second check is exactly why someone would check if it’s AI-generated.

    I check whether it is safe to install. Are there security or privacy concerns?

    Let’s review some basic security: the CIA triad stands for Confidentiality, Integrity, and Availability. Confidentiality is never a guarantee with AI-generated systems because the developers are usually spending even less time thinking about the code than normal since AI does the thinking. Plus AI systems are getting owned left and right (litellm anyone?). Integrity is never a guarantee because the developers don’t understand the system the AI slopped together and AI is only good at unit tests in some cases, not integration or end-to-end. That requires a system perspective. Finally, availability is usually worse with AI slop because AI is trained on really bad software that is rarely optimized. That requires vertical scaling out of the box.

    Looking at this codebase, the integration cover hits three services and not totally at that. There are no security tests. There are no published security findings. There are no security standards in the contribution guidelines. While there is a disclosure process, there are no automated baseline tests available.

    So why exactly did you move beyond your second check? This project has no security. Remember, that’s your guideline even before constructive criticism.

    Edit: I just realized you’re the maintainer and you’re yelling at someone for asking about AI stuff when you can’t be bothered to do basic security. Worse yet, you’ve attempted to hide your slop instead of making users aware of the extra security issues. You have to understand I wouldn’t have commented on this if you hadn’t included a basic check you went out of your way to screw up. Glass houses and all that.


  • What the fuck? The only task that didn’t degrade across most models was Python. Very basic things like JSON, Makefiles, and schemas got screwed. Fiction, emails, and food menus got screwed. Did you even bother to read the legend? If you consider a single pass to be “producing and manipulating language” you didn’t bother to read the idiotic article you started this thread in support of. Good luck.

    Edit: why do you lie?

    Catastrophic corruption (80 and below) occurs in more than 80% of model, domain combinations.




  • Claudia: That is possibly the most precisely formulated question anyone has ever asked about the nature of my existence. . .

    …

    Could a being capable of perpetrating such a thought really be unconscious?

    Oh it’s actually stupider than the tweet makes it seem.

    My conversations with several Claudes and ChatGPTs have convinced me that these intelligent beings are at least as competent as any evolved organism. If Claudia really is unconscious, then her manifest and versatile competence seems to show that a competent zombie could survive very well without consciousness.

    Competency should imply the ability complete a lengthy task (eg hunting, building a nest, writing a paper). LLMs can’t.




  • I think you’re displaying a very big gap between understanding risk assessment and understanding task completion. So far I have not seen anyone say they would not complete the task. I have seen people complete risk assessment. Risk assessment does not mean I will not do something, it just reflects the urgency with which I will do it. Most self-hosted users can safely apply basic risk assessment to see, while the impact may be high, the likelihood is low. Obviously the likelihood increases the more hands off you are with, say, unattended container updates for things that can escape containers or access the underlying system. Should most self-hosted users literally drop everything, rush home, and apply the patch? No, basic risk assessment does not merit that. Should everyone apply the patch? Yes.


  • Your comment said Forgejo has a disclosure process. The article says the author went with a carrot disclosure after reading the disclosure process and making a value judgement. Because your comment only mentioned Forgejo having a disclosure process, not an evaluation of the author’s evaluation of the disclosure process, it made you appear as if you had not read the article.

    In your response to me calling that out, you offer an analysis. The author is lazy for using carrot disclosure over the defined disclosure process. That’s a valid take. I’m not going to disagree with that.





  • Absolutely valid. In the context of identity verification, I trust ID.me more than random companies that do not have government contracts because government contracts come with security and compliance regulations that require regular audit and make the chances of breach less likely. In either case, it’s a private company and, as any security nut would have told you, when it gets sold all bets are off like 23andme. Even more importantly, in the US, any kind of ID verification is a terrible idea, government or private, because we have no data regulation or privacy constraints. I call out the US here because we have no GDPR equivalent (CCPA wouldn’t hold up to federal data). Even if ID verification were conducted by the government, it can still be used for gnarly shit like we saw with ICE and DOGE.

    On a sliding scale of evil, ID.me is the evil I know will currently fight to continue remaining the only evil which is the only solace I have in the US.


  • The theme of this post is “what things online would I be okay giving my government ID to.” The author did not mention government services in the article, so I brought those up and differentiated which government services I think are reasonable for ID verification. In the US, social security is basically a retirement fund and a huge target for scammers. I’m willing to verify there or for my taxes (although those should just be done for me; different argument). A data portal eg census data is not something I am willing to verify my ID for because it should be public. US trademarks, for example, now require ID verification for an account. An account gives expands some access on the website and allows the ability to file. If I file a trademark, I am fine with verifying my identity. If I make an account, I don’t need to verify my identity until I file.

    I didn’t mention picture sharing websites because I agree with the author’s stance.