ML researchers are not sec researchers. they need to stick to their own game.
companies need to use both camps for a good holistic view of the problem. ML is the blue team. sec researchers the red.
you survived a lot longer than I did. I feel like this for a long time now. I just use it now to do research, check HN feed for some news / procrastinating when I'm bored. Despite posting here n there I kind of assume its just talking into the wind or getting reviewed by bots.
I even like AI somewhat, some things it produces. pretty pictures i guess. scfi-fi. but still its not engaging anymore, you get saturated very quickly if things are always available.
The internet seems pretty much dead for a long long time already. a few bastions here n there of maybe-real-people talking. I had some minor hope AI/ML might actually improve things, get rid a bit of the bubbles caused by algos, but its gotten much much worse actually.
AI is not the cause of the decline or rot, but its definitely accelerating it.
A lot of things I cared about are taken over by the loud-n-stupid bunch who yells only in blanket statements and never seems to be able to produce any sound reasoning or evidence for their discourse. i call them bots despite them likely being confused humans...
The worse thing is, that it seems now more and more actual people in the real world are mimicking this behavior. Trying to say smart things about topics they know nothing about, because if chatGPT can give some smart sounding lines, why shouldnt I be able to? I am researcher of technology and the number of times people hand my vibe-coded or written totalgarbage to review or fix, (papers, experiments etc.). Their capacity to think and reason is diminishing fast. They will be fierce and toxic if you highlight this as a concern, or point at any of their hallucinations.
I've expeirenced already a few times that people, like a group of zombies, gang up on me (debate/argument) and all jump on arguments which are trivially proven to be incorrect. Even if you prove them incorrect infront of them, they will just try to eat your brain/prove you wrong by talking louder etc. - and these are 'highly educated individuals'.
Considering to leave my research job, after about 12 years of trying to work myself into such a position, and just go back where i started.. to drive a forklift. its more likely i'd be working with real humans there. and if it's a robot, atleast its a real fuckin robot, not one of these infiltrator units...
there was a guy who did relatively simple ML with RL on a druid in world of warcraft and put it into 2v2 arenas. soon enough it was completely unbeatable regardless of what human it got paired with. a 1 man army lol XD. they guy did pick the right class for it, but still. AI players done right with RL become insanely difficult to beat.
games like CS its less useful because it will be blatantly aimbotting. as it gets better it will be more and more obvious. you might be able to train it to mimick human mistakes, but i think ultimately it will be easily spotted by other players. for games without the hand-eye-coordination like turn-based games or games with 'global cooldowns' , mmorpgs etc., it will be much harder to identify.
i think normal subtle cheats like ESP when 'done right' are much more killing esports than this would.
i see a lot of complaints about certain games (windows / kernel-anti-cheat) not working. Consoles have always had exclusive titles. Windows also has them now via this anti-cheat stuff. This changes litterally nothing.
Also, you can even install windows on the box. it's one of its selling points actually... if you really want to...
kernel level anti-cheat is generally not even needed, so perhaps those companies will now consider rolling proper anti-cheat themselves rather than third-party rubbish that no one asked for.
What i also like about this console development is that it might open the door to other smaller players creating consoles in the form of mini-PC with linux and a gaming layer on there. maybe there will be (oem?)partner for valve that make more beefy machines, machines with alternate OSes (windows + skin) etc.
its a different angle that will open up many things hopefully. make it less exclusive market between essentially 3 parties.
htb isnt gonna help much -_-.
maybe to get a job, not to succeed at it...
uni isnt bad for cyber. maybe for 'red teaming' but that is only a tiny part of cyber. most companies dont even do it... (its not pentesting).
most other cyber things u can learn just fine at uni.
its the classic 'tech is behind at unis' excuse. its more important to learn the how and why than the what.
all in all it depend on where in cyber u wanna work.
if u want to do red teaming, uni is useless. HTB is a popular option but imho also useless.
you need to learn fundamental skills first. then learn to apply those within cyber.
learn to write code, debug it reverse it. basic networking. building web apps, databases. lots of enterprise systems, all flavors of virtualization, containers, etc. etc. etc.
a lot of those u can learn fine at uni!!!
'security is everywhere' and needing to know tons of stuff is more relevant for red teamers than any cyber other job.
its super helpful for blueteamers but these days your more likely just to be a security app admin than doing actual stuff to production systems to secure them -_-.
some jobs in cyber just need u to send questionaires to third parties, or use some existing platform to do basic phishing tests. its super simple stuff and htb has nothing on it...
u will need a degree to send those emails tho :p (dont ask why!)
it is used here n there but unlike bullets the attacks if they remain unknown have no armer to defend against them, but are single use.
since the 2010s atleast more than 140 countries spend over 10 mil a year on purly offensive cyber. most of those countries spend astronomical amounts more than that. that includes purchase of attack tools and exploits
opengl can do dots, lines, triangles, quads, ngons. just an option in the drawcall. pretty sure dx and vulkan will be the same but i didnt try em... what is more optimal.. idk.
maybe under the hood past the drawcall it will triangulate it. never bothered to look into that.
pretty sure the HW doesnt care one bit as it operates on vertices... not primitive shapes. (ofc like u note if ur quad is non planar it wont look like u expect but thats no concern for the driver or hw..)
this guy is unlikely to have read the games source or reversed it far enough to make these claims..what you see on the screen, how a game represents it on memory and how the translation between them works , can vary greatly.
that all being said. it is very easy to write bad code for games or use old features etc., we're all guessing unless some of you put a profiler and debugger to it... if so , show the money.
its now about how the game looks its about what features it uses to do its thing and how drivers on many platforms handle them.
this is also a reason why there are so many gfx driver patches which fix specific games...
its usually not the hardware but driver issues.
also, its infeasible for a solo dev to test on a plethora of platforms and device/driver combos, so any feedback from random internettians about performance can be really helpful to prevent breakage on platforms not available to the dev to test on.
i really love this one. its a really elegant and well informed solution. one of the nicest finds ive seen in a while was a pleasure reading how it works! thanks a lot
its super weird to me this isn't a thing, and there's resistence to the idea. I mean, if they are already masters at glacing at 100000 differnent indicators and warning messages etc. and processing them at super speeds (they really do!) then i'd say a monitor with a bunch of buttons below it to switch feeds (maybe a little more elaborate, but not tooo...) would be helpful.
the problem might be getting trained and experienced pilots to adjust to it since they are already in a certain flow of habits and skills to apply in their job, but new pilots surely could learn it as they aren't so set on their ways yet and have the opportunity to build this new data into their skillset / habits.