Normally "private-key cryptography" is the antonym of "public-key cryptography", which is to say, conventional symmetric-key cryptography such as AES or Enigma. That's what "private-key cryptography" redirects to on Wikipedia. This article is about Astra breaking a message encrypted with a private-key cryptographic algorithm, but one that was already known to be weak. I'm guessing that if you had messages encrypted with weak public-key algorithms it could break those too.
I see what you’re saying. It is generally harder to build a crypto system than to break one so I would find it hard to see potential from the design end before progress on the attack end.
> It is generally harder to build a crypto system than to break one
It was harder to design Symmetric Encryption using e.g. AES-256 than to break it? As far as I know it's design is pretty straight-forward but there is no known way to break it even with all compute of the planet at your disposal (excluding trivial ways like exposing the key).
Yes, it was. An attack only needs to work one way, but an encryption standard must resist all attacks.
There are many examples of would be crypto algorithms that died on the vine because an attack was found. It is often recommended for beginners to study breaking cryptographic algorithms long before they attempt to create them.
It didn't take 25 years to come up with Symmetric Encryption and AES but even though people try extensively to break it for the last 25 years no-one has come even close to anything that is not brute force (which is practically impossible even assuming very generous computing resources). So at least in that case coming up with the encryption seems to be a very solvable problem but breaking it seems close to impossible.
It’s worth considering AES is the result of many decades of development of prior block cipher algorithms. It is no so hard to make a variant of AES that looks similar but is not secure.
I think it’s all about the harness. If someone can find a way to create an iterative loop, with a good validator, I wouldn’t be too surprised if an LLM can end up with a solution. So far that’s how pretty much all problems have been solved by LLMs. That’s a very big if though!
Isn't the other way round?
It does not mean that breaking one is easy, but it does mean that writing one that's unbreakable it's very, very, very difficult.
For anyone wanting an introductory text for information theory & that explores some of these connections & applications, it's worth checking out the late David MacKay's 2003 textbook Information Theory, Inference & Learning Algorithms https://www.inference.org.uk/itila/
Is that why quaternions are interesting for computer graphics? I studied mathematics and computer science but never understood (or even looked so hard into) the connection.
Kind of! The scalar allows vectors and scalars to be expressed together as one object, which has a lot of computational and mathematical niceties.
The killer feature is that you can put the two together for a rotation axis (~ vector) and angle (~ scalar).
With just three gimbals (rotating circles), if gimbals A and B are aligned, you only have two degrees of freedom (rotating A is the same as rotating B). Because of this, interpolating angles is unwieldy in vector space. 'Gimbal lock' confounds animation (in hilarious but unrealistic ways) but also aerospace (four hours before 'one small step for man', just after landing, Collins joked he would like a fourth gimbal for Christmas).
I don’t think system prompts are particularly reliable way to do much at all. It’s better to put it as a hook after each response, or a skill at least so you can trigger it at will if you don’t want it everytime.
Anthropic has probably RL’d the system prompt into nothing because of their fear of the user being able to control the model. If it listened to you about the slop language, it might listen to you if you asked it to help you with no-no tasks.
The TOS argument was so awful it deserves some kind of bonus loss, but let's put that aside for now.
Let's say I paint a fence, and my neighbor complains I ruined a particular segment that shouldn't have been painted. Why can't I argue that this segment was on my property and that my neighbor gave permission to use that paint on all segments on their property?
Let's say there's a 15% chance the survey was wrong about the fence location, and a 15% chance the court disagrees with how I'm interpreting the agreement. If I can use both arguments I have a 2% chance of losing, which seems fair to me. Both arguments are reasonable, so why should I have to risk it on only one?
One fun fact I’ve heard is that the strategic oil reserve requires like 100-150 million barrels of oil to maintain operational pressure and even be functional. This means we can’t draw it down to zero; not even close.
I encourage you to do some digging using filetype:pdf on a search engine. The amount of documentation available on the SPR is fascinating.
From what I found out, turns out the question of SPR design and operational limits is a complex issue to say the least. The SPR is a big system composed of many sub-systems and there are entangled legal, policy, technical issues at play. Interesting stuff!
The US’s most important export for many years was treasury bonds. As de-dollarization occurs, the United States is bound to face some economic reckoning.
It should be noted that even still almost all US debt is held locally.
One aspect of that balanced budget that should be noted is that it does rely on a sizable funding increase from the state government.
Now, that state is legally required to have a balanced budget itself, so perhaps this is a small side note rather than a major change in perspective. I just think it’s not spoken about enough when discussing how that budget process went.
“Promise endlessly” is an oft touted criticism of social safety net programs, when the reality is that reducing vast tranches of corporate welfare is typically more than enough to solve budget shortfalls.
reply