Well yes, I’m not saying it’s a free lunch. The point is a large up-front investment buys you unlimited tokens in perpetuity for no more than the cost of electricity to power the cluster — which can be surprisingly cheap for UMA machines.
I’ve swapped off API models entirely and have been able to do a lot more useful work since I don’t have to worry about token cost or usage limits. Granted it’s slower per task (less tk/s), but overall much more productive, and private.
Perhaps they run a local model, then tokens are essentially free.
OMG. I just noticed your account name. You are hilarious. I’m an idiot. Got any lemons I could borrow?
You may want to contact a therapist. DiD is a serious condition.
Not really unique to chatbots. Here’s a couple of entries from the D.O.S. (which is apparently pay walled now? Turns out the top result is a site by a grifter completely unrelated to John Koenig).
n.
a conversation in which everyone is talking but nobody is listening— instead merely overlaying words like a spoken game of Scrabble, each player borrowing bits of others’ anecdotes to build out their own, until we reach a point when we all run out of things to say.
From anecdote, a short and often amusing account of real-life events + synecdoche, a figure of speech in which a part stands for the whole. Pronounced “uh-nek-doh-kee.”
n.
a state of exhaustion with continually being told what you want to hear, whether by marketers pandering to you or acquaintances afraid of causing offense; the longing for someone to have the heart to finally call you out on your bullshit, challenge your long-held assumptions, and push you to become a better person—which would be a far deeper kindness than trying to live and let live.
From an-, against + echoes. Pronounced “an-uh-koh-sis.”
And we’ve finally come to the point I was trying to make from the beginning. In this particular case it’s Copyright law that is bad.
For example. I have a collection of physical cookbooks that I am in the process of digitizing because I would like to have them in my self hosted Mealie server, and many of them are old and delicate so this is the only way to protect their contents.
Many of these books are rare, no longer in print, and have no digital editions. Now I have a bit more leeway in my case because the recipes themselves are not subject to copyright (only the compilation), but still the only fully legal way for me to digitize these books is to destroy them after I’ve scanned them.
This is a problem with copyright and the commercialization of knowledge in general, not something that is specific to Anthropic or any AI company. So the point is, if you want to rage out at something, make sure your ire is targeted in the correct direction.
In this case that direction should be towards reforming copyright law, and establishing broader adaptation of the copyleft philosophy (which I very much support).
And just to clarify, yes Anthropic destroying a bunch of books is less than ideal, and you can argue that the reason they want them in the first place is dumb or evil as well, but the actual act of destruction is being compelled by US law and applies to anyone in any similar situation.
Well, it’s somewhat related as it would seem that when it comes to US copyright law, making a digital copy of a book is a violation of the author’s rights unless you destroy the book after making the scan. Even then it’s still on shaky ground as it depends on what you then go and do with that digital file.
It only seems right that when you choose to cheat these hard working human authors out of their due by stealing their intellectual work, you should do so with full knowledge of its illegality.
Thanks. I appreciate you spending the time to provide me with a link. However after a close reading of the page you provided, Copyright, Fair Use, and Plagiarism for Students: “Fair Use” of Copyrighted Materials, I’ve come to the conclusion that what you have described, whether for educational purposes or otherwise, does not constitute fair use.
Generally, a fair use is any copying of copyrighted material done for a limited and “transformative” purpose, such as to comment upon, criticize, or parody a copyrighted work (Richard Stim, Stanford University Libraries).
This is news to me. Do you have any legal sources you can point me towards? I would love to maintain my own digital library of nondestructively scanned books without needing to buy them or pay any of the copyright holders.
I’m assuming there would probably still be some copyright issues. I don’t think it’s okay to just go into a library and steal books by copying them even if they aren’t destroyed right?
You got a source for this?
I don’t think I could do that at the library. Wouldn’t I need to purchase the books first?
Can you explain what happens to them?
It looks like you accidentally misspelled “using” as “burning.”
Removed by mod
I’m sorry, but it’s not my responsibility to educate you about copyright law, even though I have done what I could. I wish you the best. Good day.
When i write a sentence then do i have a copyright upon it which i can defend in law?
Umm… Yes. Yes, you do. That’s how copyright works…
The current models are all trained with data which at least some larger creator (book writer or whatever) does not want a model to be trained on.
Is this not antithetical to your stance? Or are you saying that only those entities large enough to defend themselves against the likes of OpenAI and Anthropic are deserving of being paid for use of their works?
The current models are under the current law unsustainable.
I completely agree with you here, but something like a class action lawsuit is not a viable solution.
You understand the scale right? This is a combinatorial problem: the plaintiffs in this case would be every individual and company that has ever existed, and the defendants would be not only the AI companies, but also anyone who has ever used or published AI generated content. That’s a wishful thought, not a realistic solution.
Holy shit! What an excellent idea.
I always advocate for putting AI models under UNESCO stewardship
As now will I.
That’s not how any IPO works.