I worked on a system a couple years ago with a BERT-based model (64M parameters) used for classification. The rest of the system could process data at gigabytes per second, and so here tokenization at a measly few megabytes per second really slowed things down. The model inference was more expensive than tokenization, but tokenization was still >10% of total runtime.
Logout functionality is not the only use case for token invalidation. Another significant one is to revoke a token that has been exposed, like inadvertently pushed to GitHub. In that case you want to invalidate the token to a avoid unauthorized access to your service.
With vanilla JWTs, you have no way to do this! But then if you add revocation checking on top (which people do), your JWTs are no longer stateless.
Your general point here is reasonable. But to provide some domain knowledge context: secrets are leaked _very_ often!
In public data (source code on GitHub, etc.) you can expect a prevalence somewhere in the range of 0.5-2.5 live secrets per gigabyte of content. Now yes, there are more than 8 billion people on earth now and the murder prevalence is a lot higher than 0.5-2.5 per billion. But there are _far_ more bytes of public content than there are people on earth, so in absolute terms, there are far more leaked secrets than murders.
If you look at other types of data (like internal Git forges), the prevalence is much higher.
I think you could indeed retire with $1 per leaked secret!
Until the GPU is accessible by the browser and any website can execute code on it. Or the attack can come from a different piece of software on your machine.
Rowhammer allows you to corrupt/alter memory physically adjacent to memory you have access to. It doesn't let you read the memory you're attacking.
There's PoC's of corrupting memory _that the kernel uses to decide what that process can access_ but the process can't read that memory. It only knows that the kernel says yes where it used to say no. (Assuming it doesn't crash the whole machine first)
Suppose you have access to certain memory. If you repeatedly read from that memory, can't you still corrupt/alter the physically adjacent memory you don't have access to? Does it really need to be a write operation you repeatedly perform?
> Does it really need to be a write operation you repeatedly perform?
Yes. The core of rowhammer attacks is in changing the values in RAM repeatedly, creating a magnetic field, which induces a change in the state of nearby cells of memory. Reading memory doesn't do that as far as I know.
A simple hello world in C++ can pull in dozens of megabytes of header files.
Years back I worked at a C++ shop with a big codebase (hundreds of millions of LOC when you included vendored dependencies). Compile times there were sometimes dominated by parsing speed! Now, I don't remember the exact breakdown of lexing vs parsing, but I did look at it under a profiler.
It's very easy in C++ projects to structure your code such that you inadvertently cause hundreds of megabytes of sources to be parsed by each single #include. In such a case, lexing and parsing costs can dominate build times. Precompiled headers help, but not enough...
> Now, I don't remember the exact breakdown of lexing vs parsing, but I did look at it under a profiler.
Lexing, parsing and even type checking are interleaved in most C++ compilers due to the ambiguous nature of many construct in the language.
It is very hard to profile only one of these in isolation. And even with compiler built-in instrumentation, the results are not very representative of the work done behind.
C++ compilers are amazing machines. They are blazing fast at parsing a language which is a nightmare of ambiguities. And they are like that mainly because how stupidly verbose and inefficient the C++ include system is.
> Lexing, parsing and even type checking are interleaved in most C++ compilers due to the ambiguous nature of many construct in the language.
>
> It is very hard to profile only one of these in isolation. And even with compiler built-in instrumentation, the results are not very representative of the work done behind.
Indeed, precise cost attribution is difficult or impossible due to how the nature of the language imposes structure on industrial computers. But that aside, you still end up easily with hundreds of megabytes of source to deal with in each translation unit. I have so many scars from dealing with that...
The computational complexity for lexing C++ is linear, but for parsing C++ it's super-linear, as will be many analyses. In practice, the lexing is in the noise for almost all compliers.
Good luck to naively differentiate if '>' is effectively a chevron operator or a shift operator '>>' in a random nested template expression like : 'set<vector<int>>'.
> Well, the current generation of LLMs blow away that Turing Test
Maybe a weak version of Turing's test?
Passing the stronger one (from Turing's paper "Computing Machinery and Intelligence") involves an "average interrogator" being unable to distinguish between human and computer after 5 minutes of questioning more than 70% of the time. I've not seen this result published with today's LLMs.
I only skimmed it, but I don't see anything clearly wrong about it. According to their results, GPT-4.5 with what they term a "persona" prompt does in fact pass a standard that seems to me at least a little harder than what you said - actively picks the AI as the human, which seems stricter to me than being "unable to distinguish".
It is a little surprising to me that only that one LLM actually "passed" their test, versus several others performing somewhat worse. Though it's also not clear exactly how long ago the actual tests were done - this stuff moves super fast.
I'll admit that I was not familiar with the strong version of it. But I am still surprised that nobody has done that. Has nobody even seriously attempted to see how LLMS do at that? Now I might just have to check for myself.
I would have presumed it would be a cake walk. Depending of course on exactly how we define "average interrogator". I would think if we gave a LLM enough pre-prepping to pretend it was a human, and the interrogator was not particularly familiar with ways of "jailbreaking" LLMs, they could pass the test.
A lot of AI-based PDF processing renders the PDF as images and then works directly with that, rather than extracting text from the PDF programmatically. In such systems, text that was hidden for human view would also be hidden for the machine.
Though surely some AI systems do not use PDF image rendering first!
Just thought the same and removed my edit as you comment it!
I wonder if the longer pipeline (rasterization + OCR) significantly increase the cost (processing, maintenance…). If so, some company may even remove the process knowingly (and I won’t blame them).
