Just want to correct one thing: They're not AI evangelists. They're copyright survivors from the MPAA/RIAA wars (in the 90s and early 2000s).
Plagiarism requires near-perfect copying. If you summarize, paraphrase, or re-write something using different language, you're not plagiarizing anything.
Everyone has a right to summarize or paraphrase whatever TF they want. That includes Big AI companies and individuals using LLMs.
It is not worth giving up our rights just to placate a handful of very wealthy authors.
They tested this in the research: They found that the statistical likelihood of bad tool calling was higher with watermarking compared to random seeds. At least, that was my takeaway (I read the whole thing).
This is because SynthID and similar watermarking methods for LLMs don't just change the random seed. They take additional steps (that I have yet to read about) in order to detect when someone changes a few words of the output, trying to remove the watermark.
The other takeaway is that just by knowing a watermark is being applied gives an attacker an advantage in working around safety features because then they know the output isn't based on true randomness and can take advantage of that in a similar fashion to how breaking cryptography becomes easier when the RNG isn't truly random.
I'm with you. There's so much cool stuff going on and new things to learn! Stuff to try and new tech to play with.
It sounds like the author got locked in some sort of, "I follow" loop where he got used to trusting certain companies to keep him interested and entertained, then they sold out to ads and the lowest common denominator of entertainment audiences.
The Internet is the same as it's always been! There's infinite ways to use it, full of all sorts of fun communities, new hobbies to spend too much money on (hehe), and endless amounts of people to engage with.
If your iPhone is bothering you that much, maybe it's time to try Android or go totally off the grid (so to speak) and install something like GrapheneOS on a used phone (the horror!).
How about designing your own damned phone? It's not impossible. Make a thick ass brick with a huge battery and dual screens that barely fits in your pocket but can do amazing things a normal phone can't: Give it an IR emitter so you can turn off TVs from a distance. A CO2 sensor so you can tell when any room you're in is getting stuffy. A friggin laser projector. Whatever!
Instead of being a passive intake of all things corporate Internet, make something. Contribute!
Geeze! Going to the AliExpress app to check a package is for when you remember that you ordered something over a month ago and are wondering if it'll arrive soon. Not for regular checking!
Then you look and see it made it to your country last week and give a gentle nod to the app, then close it. Because you know, "any day now, I'll have 1000 of that $0.01 thing."
All the Big AI APIs handle a million tokens of context. That's enough for 750,000 words or, several entire novels.
Context isn't an issue. It's style.
The argument the author is making is that if you use an LLM to write for you, you end up with a homogeneous work that's just like everything else.
In journalism, that's sort of what you want.
In novels/function, that's the opposite of what you want.
However, there's degrees to these things. If you're looking at a sentence you wrote, thinking there's something "off" it can be helpful to have the LLM rewrite it just to see how it thinks it could be better worded.
I've done this a few times and I do find it helpful. Just like the author, I never just accept what it gives me. I simply use it as an OK-ish copyeditor that's good at spotting issues and suggesting solutions.
Are we talking about turning everyone into philosophers and somehow ascending to a higher plane of existence? Yeah, AI isn't going to help with that (probably).
Or are we talking about useful, positive benefits to every day people like better speech recognition, tools for the visually impaired, disease research, physics research, science in general, and loads of other areas where AI is improving things?
That's what this boils down to: What are our rights?
Everyone has a right to scrape the Internet. That includes corporations who scrape the Internet to train AI models.
If we take away that right, how would the Internet even work? It wouldn't.
Example: I could tell curl right now to download this techcrunch article and all the comments about it on HN and I'd be violating no law. I'd be infringing on no one's rights.
If I then distributed these downloaded files without permission then I'd be violating copyright law. The thing it certainly would not be is theft!
People claim AI companies are "stealing" human labor but that's not true. They're saving (in their databases) the fruits of human labor and other bots/software. Then they're using that data to train AI models.
The only conclusion I can make whenever someone says "AI is theft!" is that they have no idea what they're talking about.
My assumption is that what they really mean is, "AI is bad for labor!" and possibly, "cheap AI is incompatible with capitalism." Which very well could be true.
But if AI really undermines the value of labor that much, the problem isn't the AI, it's capitalism.
> People claim AI companies are "stealing" human labor but that's not true. They're saving (in their databases) the fruits of human labor and other bots/software. Then they're using that data to train AI models.
And profiting on it on a scale that’s hard to fathom. Someone who spent effort creating a great resource or doing some research and maybe got some income via donations, ads, whatever. Now that information from their resource is distilled into a big model. The original author is screwed, the model provider makes money through the effort of everyone else. It worked well for everyone before, because there was recognition, prestige, a sense of doing good for people, even a chance for some income. That’s completely eliminated with AI.
Good luck convincing the current DOJ to do anything useful at all though! It is currently intentionally stacked with incompetent cronies who have been told that their job is to attack the President's enemies and ignore the misdeeds of his allies.
It will remain like that until he's gone (and not replaced with another Republican wannabe dictator).
It will remain this way in slightly differnt shades of colors until there is a systemic change. IM sorry this isn't a one party problem. There are people across the isle that are allowing this to continue.
Im not in support of any party btw. Im only in support of humanity doing humane things.
The grandparent comment said "liable". That's civil court, not criminal, and doesn't need the DOJ to be involved. HuggingFace and/or RubyGems could sue OpenAI.
States also have their own laws against unauthorized computer use (hacking). A state Attorney General could bring a suit under those laws, regardless of who is in the white house.
Are you kidding me? It costs a fraction of the competition and you can use just about any pin for anything you want! It makes designing boards for the RP series so simple and easy!
Contrast that with STM32 where you have to constantly reference the data sheet to make sure you're using the correct pins for say, I2C when you're running in a particular mode. And since you're running in that mode, you now can't use other pins for certain things. Ugh! I hate it!
Then there's the issues with picking the right-size part number and making sure the pins you want are 5V tolerant, checking that you're using the correct ADC, and... Ugh! It's a pain in the ass!
Other chips aren't much better as they all have those same problems.
ESP32 is nice because of the zillions of built-in peripherals but configuring the damned thing can get really complicated really fast. It's so complicated, in fact, that they had to make a Makefile (as in `make menuconfig`) that rivals the Linux kernel in the sheer number of options/submenus.
nRF5x also has completely configurable pins, and if you're worrying about awkward pinouts making PCB design slightly more annoying then you're definitely not in a cost sensitive project.
Plagiarism requires near-perfect copying. If you summarize, paraphrase, or re-write something using different language, you're not plagiarizing anything.
Everyone has a right to summarize or paraphrase whatever TF they want. That includes Big AI companies and individuals using LLMs.
It is not worth giving up our rights just to placate a handful of very wealthy authors.