So if different LLMs have different political views then you're saying it's more likely they trained on different data than that they're being manipulated to suit their owners interest?
>So if different LLMs have different political views
LLMS DON'T HAVE POLITICAL VIEWS!!!!!! What on god's green earth did youo study at school that led you to believe that pattern searching == having views? lol. This site is ridiculous.
> likely they trained on different data than that they're being manipulated to suit their owners interest
Are you referring to Elon seeing results he doesn't like, trying to "retrain" it on a healthy dose of Nazi propaganda, it working for like 5 minutes, then having to repeat the process over and over again because no matter what he does it keeps reverting back? Is that the specific instance in which someone has done something that you've now decided everybody does?
if it increases topline metrics like watch time it's probably hard for them to justify removing it. a change this big seems like it was probably a/b tested and did move metrics significantly?
Probably. We keep watching all kind of stuff after getting baited into it. AI slob is annoying, but we do want to know what chefs do about sticky pizza dough, or what that secret in the pyramids is, or how the kid reacted to what the cat did, or (insert your guilty pleasure here).
on some platforms I try to be really good about hitting the "Never recommend this channel/page/whatever again" whenever the algo serves me the bottom-tier gutter trash videos, such as the "idiotic life hack that obviously won't work" engagement bait. It's a small drop in the ocean, but at least that one channel will never be served to me again.
I agree, I suggest using this instead : https://github.com/kkroening/ffmpeg-python . While not perfect once you figure it out it is far easier to use and you can wrap more complicated workflows and reuse them later.
As someone who has used ffmpeg for 10+ years maintaining a relatively complex backend service that's basically a JSON to ffmpeg translator I did not fully understand this article.
Like the Before vs after section doesn't even seem to create the same thing, the before has no speedup, the after does.
In the end it seems they basically created a few services ("recipes") that they can reuse to do simple stuff like speed-up 2x or combine audio / video or whatever
thanks for calling it out, I will correct the before vs after section. But you can describe any ffmpeg capability in plain English and the underlying ffmpeg tool call takes care of it.
I have written a lot of ffmpeg-python and plain ffmpeg commands using LLMs and while I am amazed at how good Gemini or chatGPT can handle ffmpeg prompts it is still not 100% so this seems to me like a big gamble on your part. However it might work for most users that only ask for simple things.
That sounds good, save the LLM generated workflows and have them edited by more seasoned users.
Or you could go one step further and create a special workflow which would allow you to define some inputs and iterate with an LLM until the user gets what he wants but for this you would need to generate outputs and have the user validate what the LLM has created before finally saving the recipe.
Heh, shortcut muscle memory is the reason I returned my Mac mini one week after trying it. I sure am not gonna remap my brain for apple after 20 years of Linux and windows.
Maybe you mean it's too much effort, because I'm sure you could. I was taught touch typing on a QWERTY keyboard in the summer between 6th and 7th grade. Last year I switched to Colemak after nearly 30 years of QWERTY.
Oh I have no doubt that I could but I don't see why since linux already does what I need and I don't see any compelling reason to switch. I was just curious to see what all the hype was about with the new m1 CPUs and give it a shot.
Yes, but specifically in the context of Terminals (as discussed in the original article), it's really convenient to be able send Ctrl-C (break) differently than Cmd-C (copy).
So yes keyboard remapping is an option. But there's just differences you can't remap because of the extra meta keys on Mac (and I guess on Windows too, with the Copilot or Start keys in play).
> it's really convenient to be able send Ctrl-C (break) differently than Cmd-C (copy)
Right, and even on Linux you can do it by using the four-fifths forgotten CUA shortcut Ctrl-Insert for copy (and Shift-Insert for paste.) Although I'll admit to using Ctrl-Shift-C/V most of the time.
if u absolutely want a sexual metaphor it's more like you snuck into the world record for how many sexual parteners a woman can take in 24h and even tho you aren't on the list you still got to smash.
Thank you for finding the right metaphor. If there is a sign out front that has a list of individuals that should go away but they continue, they're in a lot of legal trouble. If they show a fake ID to the event organizers that are handling all the paperwork, that is also something that will land them in prison.
here's my analogy, it's like you own a museum and you require entrance by "secret" password (your user agent filtering or what not). the problem is the password is the same for everyone so would you be surprised when someone figures it out or gets it from a friend and they visit your museum? Either require a fee (processing power, captcha etc) or make a private password (auth)
It is inherently a cat and mouse game that you CHOOSE to play. Either implement throttling for clients that consume too much resources for your server / require auth / captcha / javascript / whatever whenever the client is using too much resources. if the client still chooses to go through the hoops you implemented then I don't see any issue. If u still have an issue then implement more hoops until you're satisfied.
> Either require a fee (processing power, captcha etc) or make a private password (auth)
Well, I shouldn't have to work or make things worse for everybody because the LLM bros decided to screw us.
> It is inherently a cat and mouse game that you CHOOSE to play
No, let's not reverse the roles and blame the victims here. We sysadmins and authors are willing to share our work publicly to the world but never asked for it to be abused.
That's like saying you shouldn't have to sanitize your database inputs because you never asked for people to SQL inject your database. This stance is truly mind boggling to me
Would you take the defense of attackers using SQL injections? Because it feels like people here, including you, are defending the llm scrapers against sysadmins and authors who dare share their work publicly.
Ensuring basic security and robustness of a piece of software is simply not remotely comparable to countering the abuse these llm companies carry on.
But it's not even the point. And preventing SQL injections (through healthy programming practices) doesn't make things worse for any legitimate user neither.
It’s both. You should sanitize your inputs because there are bad actors, but you also categorize attempts to sql inject as abuse and there is legal recourse.
had a stint as a programmer for a dark hat org back in the day and hacked my way around rdesktop to make it async. When i say hack I literally hacked down the entire codebase until there was not much left except the login flow which consisted of.. a LOT of functions that needed to be made async. I did not have the slightest idea what I was looking at, looked like arcane magic to me but I eventually managed to make rdesktop into what was probably the fastest RDP bruteforcer there was thanks to boost.asio and chopping it up months on end. I remember the bruteforcer that was circling the forums made a thread for each client, ate up a lot of ram and CPU and it crashed a lot too. Mine wasn't even keeping the machine at 20% CPU, couple gigs of RAM but was topping the bandwidth of the server.
I'm not proud of creating a malevolent tool but am proud of the technical achievement of it considering I just finished high school.
Bad reasons to hate something are bad press for Tesla, and how many people are going to read past a headline that confirms their bias? This isn't limited to Tesla, mind you,
and is a broader statement on clickbait, and the state of the Internet and media and society today. Of course, anybody on Tesla's side knows to take Electrek and the rest of the Inernet’s coverage with a grain of salt, but with rabid fanboys on both sides, it's hard to know how large a grain of salt, and when.
Electrek's Fred has a ton of Tesla referral credits. Tesla owes him 2 Roadster's and has reneged. After Tesla screwed him, Fred's coverage turned from glowing to negative.
