• 0 Posts
  • 34 Comments
Joined 1 year ago
cake
Cake day: August 12th, 2024

help-circle


  • Terminology: revoked means the issuer of the certificate has decided that the certificate should not be trusted anymore even though it is still valid.

    If a attacker gets access to a certificates key, they can impersonate the server until the validity period of the cert runs out or it is revoked by the CA. However … revocation doesn’t work. The revocation lists arent checked by most clients so a stolen cert will be accepted potentially for a very long time.

    The second argument for shorter certs is adoption of new technology so certs with bad cryptographic algorithms are circled out quicker.

    And third argument is: if the validity is so short you don’t want to change the certs manually and automate the process, you can never forget and let your certs expire.

    We will probably get to a point of single day certs or even one cert per connection eventually and every step will be saver than before (until we get to single use certs which will probably fuck over privacy)



  • You don’t need something ever. Sometimes you just want something because the alternative is realy bad. I don’t need to eat. I want to eat because I don’t want to starve.

    I want to watch a movie with my partner at the agrees time because otherwise they will be mad. I want to access my digitalized documents to send a letter in time because otherwise I will have to pay late fees. I want to access my gameserver because that’s the one time a week I get to have fun with my friends from my college time.

    There are many situations where I’d rather do the thing I want instead of doing maintenance.





  • You (and anyone reading) should replace 2048 with “threes”. Its the original game that 2048 is based on (developer claims they never knew threes already existed but not sure about that), it has way more details and love aaaaaand it actually works as a game. In 2048 you win by going down,left,down,left,down,left, repeat until high score. In threes you actually have to think



  • That is the smallest scale of self hosting. The server and the client are the same device. It is also the most insecure way as you probably don’t have any backups and very limited storage space.

    Actually self hosting is the next step when you decide you want 5+ TB of data and have it automatically create backups. Digital storage media degrade pretty quickly and if you just have your movies on a hard drive in your computer, after 5-10 years you might start to lose quality or some files completely.




  • Sometimes I wonder how people can be so naive. Even if they don’t know just how much money is thrown at AI, just the fact that it is now in literally anything and on the news every day should be a clue that it is a huge industry. You wouldn’t expect the guy that changes your cars tires to be able to single handedly build a car that is better, faster, cheaper and more efficient than all car companies together.


  • groet@feddit.orgtoMemes@lemmy.ml3 trillion dead
    link
    fedilink
    arrow-up
    1
    arrow-down
    6
    ·
    2 months ago

    The third Reich was not in the name of God. The Church had to subjugate under the Nazis. Same with Stalin and the communists. They hated religion and I am pretty sure they committed quite a few atrocities in the USSR, Mao in China, Pol Pot in Cambodia …

    God is a convenient reason for atrocities but far from the only/most prevalent one.




  • Anyone with a stake in the development of AI is lying to you about how good models are and how soon they will be able to do X.

    They have to be lying because the truth is that LLMs are terrible. They can’t reason at all. When they perform well on benchmarks its because every benchmark contains questions that are in the LLMs training data. If you burn trillions of dollars and have nothing to show, you lie so people keep giving you money.

    https://arxiv.org/html/2502.14318

    However, the extent of this progress is frequently exaggerated based on appeals to rapid increases in performance on various benchmarks. I have argued that these benchmarks are of limited value for measuring LLM progress because of problems of models being over-fit to the benchmarks, lack real-world relevance of test items, and inadequate validation for whether the benchmarks predict general cognitive performance. Conversely, evidence from adversarial tasks and interpretability research indicates that LLMs consistently fail to learn the underlying structure of the tasks they are trained on, instead relying on complex statistical associations and heuristics which enable good performance on test benchmarks but generalise poorly to many real-world tasks.