yorwba

34 ms·

691.
by yorwba 1y ago
Did you forget about the 𓈐 or does it get merged into the preceding word?
692.
by yorwba 1y ago
(I'm not an egyptologist.) The Thesaurus Linguae Aegyptiae has a breakdown of 𓅓𓂋𓄋𓏏𓏝𓊹𓊵𓏏𓅓𓇾𓂋𓇦𓂋𓆑 / jm.j-rʾ-wpw.wt-ḥtp.w-nṯr-m-tꜣ-r-ḏr=f / "overseer of apportionments of the god's offering(s) in the entire land"
693.
by yorwba 1y ago
In section 5.7.5, they fine-tune for "11 low-resource languages, with between 5-10 hours of training data and at least 1 hour of validation splits." "CTC fine-tuning takes ≈1 hour of walltime on 32 GPUs for the 300M scale.&qu
694.
by yorwba 1y ago
The Ethnologue link in footnote 7 of the paper has utm_source=chatgpt.com at the end, so I suspect whoever was tasked with listing languages and determining their status thought this wasn't important enough to do it themselves and ju
695.
by yorwba 1y ago
"Does that statement actually reflect what Netflix truly think and that they actually believe GenAI shouldn't be used to replace or generate new talent performances?" The if-statement "If you want to do X, you need to ge
696.
by yorwba 1y ago
The part you quote is part of the list of conditions for an if-statement, so how could it be a lie?
697.
by yorwba 1y ago
"If you can confidently say "yes" to all the above, socializing the intended use with your Netflix contact may be sufficient. If you answer “no” or “unsure” to any of these principles, escalate to your Netflix contact for mor
698.
by yorwba 1y ago
Their company website also has shorter descriptions of the scheme for people with varying amounts of background knowledge: https://raveltech.io/ The implementation doesn't appear to be public, but their GitHub strongly
699.
by yorwba 1y ago
The transcript in question may or may not be this one: https://newmitbbs.com/blog/articles/9694
700.
by yorwba 1y ago
Synthetic data doesn't have to come from an LLM. And that paper only showed that if you train on a random sample from an LLM, the resulting second LLM is a worse model of the distribution that the first LLM was trained on. When peopl
701.
A Week of Teaching(fivetwelvethirteen.substack.com)
2 points by yorwba 1y ago|0 comments
702.
by yorwba 1y ago
Currently on the front page: https://news.ycombinator.com/item?id=45833162
703.
by yorwba 1y ago
Even when you do get a sync conflict, Syncthing will rename one of the copies and then you can have KeePassXC merge the two files back into one. So that's still pretty much hassle-free.
704.
by yorwba 1y ago
It's true that to train more information into the model you need more trainable parameters, but when people ask for small models, they usually mean models that run at acceptable speeds on their hardware. Techniques like mixture-of-expe
705.
by yorwba 1y ago
The US lost in the gambling case because their restrictions on foreign websites were stricter than those on domestic ones. The GATS doesn't prohibit countries from regulating trade, they only have to do so in a non-discriminatory manne
706.
by yorwba 1y ago
It's called checked luggage for a reason. They already check every bag, so checking for one more thing is hardly a problem.
707.
by yorwba 1y ago
The "attention is all you need" paper did not invent attention mechanisms. It showed that existing models that were already using attention could have their non-attention parts removed and still worked. So those other parts were u
708.
by yorwba 1y ago
Most libraries probably don't stock many books like that. They'd just waste shelf space until they get discarded in the end.
709.
by yorwba 1y ago
It is a compiler. It is not a compiler for Python, because there are valid Python programs it can't compile and isn't intended to compile.
710.
by yorwba 1y ago
The memory is model-readable but not model-writable, so you still need to train via backprop to get the memory to store useful data.
711.
by yorwba 1y ago
That test page doesn't seem to use any features current Chrome doesn't support. Or do you just mean that the appearance isn't identical to the TeX rendering even if you use a font like Latin Modern?
712.
by yorwba 1y ago
One problem with testing one change at a time is that if you can only run a small number of experiments because each one requires many GPU hours to get results, you can also only test a small number of changes. If you can come up with and i
713.
by yorwba 1y ago
If there's a default (I don't think there necessarily has to be one) there has to be somebody who decides what the default is. If most people trust them, that person is either very trustworthy or people just don't care very m
714.
by yorwba 1y ago
Sybil attacks are a problem when you care about global properties of permissionless networks. If you only care about local properties in a subnetwork where you hand-pick the nodes, the problem goes away. I.e. you can't use such a schem
715.
by yorwba 1y ago
Unless you can be fooled into trusting a fake endorser, that island might just as well not exist.
716.
by yorwba 1y ago
I assume you're talking about this? " Half the mean width of a polyhedron P is equal to the expected value of max θ^T x subject to x ∈ P, where θ ∈ S^(d−1) is uniformly random distributed with respect to the Haar measur
717.
by yorwba 1y ago
Importantly, they're talking about continuous representations, i.e. the output logits. For there to be a loss, you'd need two different tokens to produce the exact same logits, which is even less plausible. But as soon as you sa
718.
by yorwba 1y ago
I wouldn't bet that every smart TV Crunchyroll wants to be available on has more processing power than your phone from 2015 (some of those TVs might be older than that), but yes, it's probably less about hardware capabilities than
719.
by yorwba 1y ago
Rocky Linux has definitely had multiple releases: https://rockylinux.org/ I'm not sure what that Vercel site is, maybe a supply-chain attack of some kind.
720.
by yorwba 1y ago
For people who couldn't write that article themselves, but would like to read it if it existed. Same as big Wikipedias, just on a smaller scale.

More ›