We need open distributed "p2p" models a la bittorrent , that allow individuals to share their computer power for inference. So that the models cant be censored and everyone can run SOTA models.
It doesn't have to be free, we have the means to transact in a p2p fashion electronically as well.
This, and distributed LLM inference. We are at a point where no single person can setup a rig to run a SOTA model, it is just too expensive.
So we must build and adopt frameworks that allow individuals to share resources to run SOTA models in a distributed manner. That way they will also be non-censorable by governments.
Also The only way to prevent that one entity weaponizes it, is by giving EVERYONE access to it.
Perhaps some radical MoE where you download _exactly_ the components you need as you need them. Currently MoE is switched usually on per-token per-layer basis, so you need all weights locally. But e.g. Apple made one which pre-selects all experts based on prompt embedding. That might be further scaled up - e.g. predict exactly what's needed
> The only way to prevent that one entity weaponizes it, is by giving EVERYONE access to it
There is a middle way; the policy space also includes government regulating both access and monopoly.
I’m opposed to monopolies of this tech, but I hope the risks of giving everyone jailbroken AGI/ASI are clear.
As a toy example you could imagine a Universal Basic AI where government subcontracts to (n_quorum) labs, everyone gets a token budget, but operating the APIs comes with the safety controls.
If everyone does get to run their own jailbroken AGI, then the only stable societal norm I see is A LOT of surveillance to make sure nobody is building CBRNE threats. This doesn’t seem like a clear win from a civil liberty perspective, though I could see the argument.
Likewise, but for different reasons: We had a [pirated] copy of the game, and thus didn't have the manual.
Down here in my city in Mexico that's basically how everyone played it, so most of us played only the first level.
At some point, I was tinkering with the "x tree gold" program and saw the "hex view" thing. I remember opening Prince's .sav file, which was a very small file that only appeared after you saved. After tinkering with the numbers I managed to appear in the next level .
It was my 5 minutes of fame at my computer class when I arrived and showed that I had passed the bottles room.
And I became fascinated by cracking at that point.
I probably will be burned for this, but with the help of an LLM I wrote a tiny program that captures video from a browser screen (Xbox live online FPS game), passes the video images through a small trained NN that recognizes people forms and presents the video on another screen. That way I can place a green overlay on enemies and they are easier to see on PVP matches.
All that in around 100 lines of code, including the training/fine-tuning of the tiny YOLO nn.
That's not how that term works. Input latency/lag refers to the time differential between triggering some kind of input and the action of that input showing up on your screen.
you described end-to-end latency, input latency + processing latency + output latency.
input latency is a measure of latency between the human input and the computer receiving the input.
output latency measures the time between a computer commanding a certain pixel to change color on the screen and the color change actually taking place.
It's not my fault these terms are often used incorrectly.
Most of the time, what you see when you see someone test "input latency" is that person actually testing end-to-end latency, which is input latency + processing latency + output latency, as it is difficult (but not impossible) to test only one of these without special hardware. testing all three at once is easy.
A proper input latency test would be (for example) some external tool sending keypresses to a computer and measuring (via a hardware debugger or some other hardware-level tooling) how long it takes for the program you are interested in to receive that input.
As stated previously, output latency is the time between your program commanding something on the screen to change and that change actually happening.
there's a third latency in this stack, and that's your program itself. how long between the time it has received an input before it commands an output device to change its output. processing latency.
for the purposes of end-to-end latency testing hardware, the processing latency is effectively zero.
all three of those stack up to become "end-to-end latency" which is what most tooling available to end users measures.
Why would you get burned for this? For people who don't know, sure, this smells of 'cheating' but it's essentially just an exercise in computer vision. Also, I'd posit that using various DMA type tools in the video game modding arena are impressive and unusually secretive because they make money on subscription services. It's fascinating to me how well some of those tools and mux hardware works. When you combine that with hardware we can now buy like the DGA with more onboard memory and faster interconnects, it's going to make reverse engineering, malware analysis, forensics -- a very interesting time for those who know what they are doing.
Would you either share the code, or describe the approach? I’ve asked LLMs how to do something similar a few times recently, and received several different potential approaches. I’m still no clearer which route to take!
My wife took on building the plant inspired lego sets several months ago. They are OK in that they don't look that ugly scattered around the house. I like it because it is a hobby she enjoys a lot and it makes it easier for me to buy "the right" gifts haha.
I've never had the patience to build those. I think I have PTSD from my childhood, when my dat bought us a "cheap" brand of lego-like toys (called TENTE I think) for which the bottom pieces fell as you plugged the top pieces.
They were a race of humans that hated contact with other people. Each of of them lived in estates separated by acres of space.
We keep pushing our culture/society towards that sort of thing. We keep writing into to this "social" media (including what I am just writing) which is not social at all (but more akin to shouting opinions the middle of a mall).
Someone way more eloquent than me should write a column titled "Why do we read?"
Way back in the past (around 30 years ago) I remember reading an article on "how to read a book" or a similar subject. They argued that, you should not skip the acknowledgments, preface and other "personal" related sections of a book, because it was there where you got a glimpse of the person that was writing the book. The idea being that, you should had in mind that the person writing was explaining something through you.
Carl Sagan even has a video where he argues Books/Writing is some sort of communication through time.
Now, this has been the case historically: A person writes some text (even in botched language like my writing, as English is not my first language) with thinking that someone else in the future will read the ideas and reason about them.
But what about text written by an LLM? Does it have inherent intention? When reading LLM text, it feels like looking at those "this is not a person" photos. Yeah, they are words, yeah they form sentences and paragraphs but... they lack "soul".
It's not "Why do we read?" but something related that is coming up a lot in my thinking lately is Walter J. Ong's "Writing is a Technology that Restructures Human Thought".
Isn’t “Writing is a Technology that Restructures Human Thought” another way of saying that “feedback has an effect”?
If so, this seems to be a trivial (still worthy) assertion.
For example, I intend to, say, construct a shed. I make mistakes that I only see because I actually constructed. I revise future endeavours involving sheds.
I admit to not having read this piece, and am merely reacting to the title.
—-
Okay, I got through the first paragraph of Walter’s writings. While I nod to the bitterness (I assent to the existence of it), I do not bow.
Not normally, no. Can you point to a divergence of the bitterness in the subsequent text?
What I find to be the normal pattern (by intuition) is that the condensed leading text belies the expansive following text. This is likely lazy (a shortcut) and I am open to correction at your effort. If a call to your effort (I apologize) is unpalatable then I concede.
> Way back in the past (around 30 years ago) I remember reading an article on "how to read a book" or a similar subject. They argued that, you should not skip the acknowledgments, preface and other "personal" related sections of a book, because it was there where you got a glimpse of the person that was writing the book. The idea being that, you should had in mind that the person writing was explaining something through you
Maybe? That is one reason to read, but there are a lot of other reasons, too. It doesn't mean you are doing it wrong if you want to read something and don't care at all about the person who wrote it.
Yeah, but when we talk about food, there are different tastes, and there is stuff like "you can also use it as a doorstop". Fine, but that doesn't make a doorstop food.
> people had to wait overnight to continue vibe coding because vendors blocked further API calls for many hours at a time
Tangential but this is funny. Back in the early 90s, I did a lot of BASIC programming in the family computer, this was before we had Internet. I could spend hours.and hours in front of the computer doing stuff.
Fast forward to around 2010 I remember a distinct feeling one time the internet went off at home. Sitting in front of the computer and feeling that it was "useless" because it wasn't connected to the net.
We are getting to that point in coding apparently: 5-10 years ago, everyone programmed just by typing commands, looking at S.O. and thinking. Now, if we open our "IDE" and it doesn't have access to The Brain, we are left just standing there looking in awe at the machine.
dunno, I have electricity problems (especially on winters when Russia strikes the hardest on infra) but I usually have this time as a downtime for lightweight C coding in Termux and retro gaming, all on Galaxy Note 8 (Android 9!!) + power bank.
I guess it feels less like a problem when you have that problem regularly and are forced to adapt. and I guess I'll just HAVE to switch to Pixel 10 when Pixel 11 comes out - the integrated Linux terminal right there is awesome. or maybe just get a MacBook like most around me did
I keep going back to Sublime Text when everything in VS Code becomes too much. Last time I looked at Sublime, I was like “Damn, the last update was from 2024? Must be dead.” Until I realized the lack of updates was because it was fully functional for what they wanted as is without connecting to the internet at all.
It isn't just a psychological feeling though. We've (unnecessarily) offloaded everything to the net, so there's a very real element of uselessness that kicks in when there's no connection.
E.g. back when you were coding BASIC, you probably had magazines and either ended up copying a lot of code by hand, or if you were lucky the mag came with a floppy disk. Now no such magazines exist. Manpages were all local, now it's readthedocs online. Fat local-friendly standard libraries in almost all languages have been modularised and package managers for the most part expect to install stuff by fetching it from the net.
So unless you have heavily prepared for the cyberapocalypse or sth, there really is not much you can do on your machine when the internet goes down.
On the other hand, however, when you can prepare in advance, it's great to shut off the net for a while. I do my most productive coding during flights, for example.
Yeah, I also feel that this negative aspect of "AI" adoption is not much discussed overall - massive centralization and dependency on a remote service woth something as important as computer programming.
It doesn't have to be free, we have the means to transact in a p2p fashion electronically as well.
reply