755
[Kevin Comics] Zuckerberg AI
(sh.itjust.works)
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
this post was submitted on 22 Nov 2024
755 points (98.1% liked)
Comic Strips
25930 readers
3763 users here now
Comic Strips is a community for those who love comic stories.
Rules
-
๐ Be Nice!
- Treat others with respect and dignity. Friendly banter is okay, as long as it is mutual; keyword: friendly.
-
๐๏ธ Community Standards
- Comics should be a full story, from start to finish, in one post.
- Posts should be safe and enjoyable by the majority of community members, both here on lemmy.world and other instances.
- Any comic that would qualify as raunchy, lewd, or otherwise draw unwanted attention by nosy coworkers, spouses, or family members should be tagged as NSFW.
- Moderators have final say on what and what does not qualify as appropriate. Use common sense, and if need be, err on the side of caution.
-
๐งฌ Keep it Real
- Comics should be made and posted by real human beans, not by automated means like bots or AI. This is not the community for that sort of thing.
-
๐ฝ๏ธ Credit Where Credit is Due
- Comics should include the original attribution to the artist(s) involved, and be unmodified. Bonus points if you include a link back to their website. When in doubt, use a reverse image search to try to find the original version. Repeat offenders will have their posts removed, be temporarily banned from posting, or if all else fails, be permanently banned from posting.
- Attributions include, but are not limited to, watermarks, links, or other text or imagery that artists add to their comics to use for identification purposes. If you find a comic without any such markings, it would be a good idea to see if you can find an original version. If one cannot be found, say so and ask the community for help!
-
๐ Post Formatting
- Post an image, gallery, or link to a specific comic hosted on another site; e.g., the author's website.
- Meta posts about the community should be tagged with [Meta] either at the beginning or the end of the post title.
- When linking to a comic hosted on another site, ensure the link is to the comic itself and not just to the website; e.g.,
โ Correct: https://xkcd.com/386/
โ Incorrect: https://xkcd.com/
-
๐ฌ Post Frequency/SPAM
- Each user (regardless of instance) may post up to five (5 ๐) comics a day. This can be any combination of personal comics you have written yourself, or other author's comics. Any comics exceeding five (5 ๐) will be removed.
-
๐ดโโ ๏ธ Internationalization (i18n)
- Non-English posts are welcome. Please tag the post title with the original language, and include an English translation in the body of the post; e.g.,
Sรญ, por favor [Spanish/Espaรฑol]
- Non-English posts are welcome. Please tag the post title with the original language, and include an English translation in the body of the post; e.g.,
-
๐ฟ Moderation
- We are human, just like most everybody else on Lemmy. If you feel a moderation decision was made in error, you are welcome to reach out to anybody on the moderation team for clarification. Keep in mind that moderation decisions may be final.
- When reporting posts and/or comments, quote which rule is being broken, and why you feel it broke the rules.
Banned Artists
The following artists are banned from the community.
- Jago
- Stonetoss
- GPrime85
It should be noted that when you make reports, it is your responsibility to provide rational reasoning why something should be removed. Saying it simply breaks community rules is not always good enough.
Web Accessibility
Note: This is not a rule, but a helpful suggestion.
When posting images, you should strive to add alt-text for screen readers to use to describe the image you're posting:
Another helpful thing to do is to provide a transcription of the text in your images, as well as brief descriptions of what's going on. (example)
Web of Links
- !linuxmemes@lemmy.world: "I use Arch btw"
- !memes@lemmy.world: memes (you don't say!)
Other Comic Communities of Interest
- !bloomcounty@sopuli.xyz
- !calvinandhobbes@lemmy.world
- !cyanideandhappiness@discuss.online
- !exo@discuss.online
- !foxtrot@slrpnk.net
- !garfield@lemmy.world
- !moomin@sopuli.xyz
- !oglaf@discuss.online (NSFW)
- !outland@slrpnk.net
- !pbf@discuss.online
- !peanuts@discuss.online
- !smbc@discuss.online
- !theboondocks@slrpnk.net
- !thefarside@sh.itjust.works
- !xkcd@lemmy.world
- !NoseEars@lemy.lol
founded 3 years ago
MODERATORS
Only the inference code of LLaMA (which runs the model) is open-source. The model itself is not, as you're given neither the training data, nor the model weights.
I don't know much about AI models, but that's still more than other vendors are giving away, right? Especially "Open"AI. A lot of people just care if they can use the model for free.
How useful would the training data be? Training of the largest Llama model was done on a cluster of over 100,000 Nvidia H100s so I'm not sure how many people would want to repeat that.
scientific institutions and governments could rent enough GPUs to train their own models, with potentially public funding and public accountability, and also itโd be nice to know if the data llama was trained with was literally just facebook user data. iโm not really in the camp of "if user content is on my site then the content belongs to me".
Without the same training data you wouldn't be able to recreate the results even when having the computing power. Thus it's not fully open source. Training data is a part of the source to create the result, "LLM". It's like having to add your own lines of code to open source program to make it work because the company doesn't provide it.
It turns out these clusters are being used very inefficiently, seeing how Qwen 2.5 was trained with a fraction of the GPUs and is clobbering models from much larger clusters.
One could say Facebook, OpenAI, X and such are "hoarding" H100s but are not pressured to utilize them efficiently since they are so GPU unconstrained.
Google is an interesting case, as Gemini is getting better quickly, but they presumably use much more efficient/cheap TPUs to train.
Open datasets are getting much better (Tulu for an instruct database/recipe is a great example), but its clear the giants still have "secret sauce" that gives them at least a small edge over open datasets.
There actually seems to be some vindication of using massively multilingual datasets as well, as the hybrid chinese/english models are turning out very good.
How on earth would you distribute the model for inference without the weights? The gradients are obviously gone so you can't continue training on the model. Maybe you can still do some kind of LORA?