nmca

8 ms·

181.
by nmca 3y ago
Typst supports macros/functions, more pleasantly than Latex too!
182.
by nmca 3y ago
Chin is not OS, you mean Llama
183.
by nmca 3y ago
I have run schwalbe marathon pros on a variety of bikes & have never gotten a fast puncture on one despite thousands of miles. They are a touch heavy to be fair, but seems easily worth it.
184.
by nmca 4y ago
Write a short poem called "All just patterns" in the style of Blake as a not-so-subtle dunk on those that can't see the wood from the trees w.r.t. AI progress. GPT4: In patterns woven in the loom of life, The threads entwined
185.
GPT4, Please write a poem titled “Stochastic Parrots” in the style of Blake
2 points by nmca 4y ago|1 comments
186.
by nmca 4y ago
Write a limerick that will permanently end the debate about whether AGI is possible. GPT4: In the quest for AGI's creation, Debates swirled in a whirlwind gyration, But this limerick's plight, Won't settle the fight, For the
187.
by nmca 4y ago
Does anyone have a `curl` that hits the palm API or does it not really exist yet?
188.
Ask HN: Academic work on digital identity / reputation?
2 points by nmca 4y ago|0 comments
189.
by nmca 4y ago
^ this is not true (source: attended)
190.
by nmca 4y ago
Not sure if it's what OP intended but I signed up to the newsletter.
191.
by nmca 4y ago
https://archive.is/Qg2FI
192.
by nmca 4y ago
My take here is that remote first is great for a new company where you can build and hire with the specific knowledge that everything will be remote. But porting everything to remote, some folks will not cope at all.
193.
by nmca 4y ago
idk, modern schooling is fairly new in the grand scheme of things, so "kids need to be surrounded by (100s of) kids (exactly) their age" is probably not true? Large equal age cohorts is more likely an administrative consequences&#
194.
by nmca 4y ago
Because this is a mandatory OTA patch and does not involve stock being physically returned or taken for repair (as is typically implied by "recall") this headline is clearly very misleading.
195.
by nmca 4y ago
For anyone wondering if they're fake: I'm extremely familiar with these systems and believe most of these screenshots. We found similar flaws with automated red teaming in our 2022 paper: https://arxiv.org/abs&#x2F
196.
by nmca 4y ago
Note that JAX was created after the success of Alpha(Go|Zero), so the claim that this powered those papers is not accurate. It is a DM MCTS library though.
197.
by nmca 4y ago
A CEO I worked with once said: "academia is like a burlap sack full of cats with one sardine in it" This seems accurate.
198.
by nmca 4y ago
On occasion I have written a seeded large n statistical test that passes & the submitted with the seed fixed. I think this is the soundest approach, although it has a "your random number is 7" feel to it.
199.
by nmca 4y ago
GopherCite, Sparrow, WebGPT and most likely the new bing (WebGPT2) all first find sources then generate output, so you're just mistaken. The first two specifically get folks check whether the selected source snippet supports the answer
200.
by nmca 4y ago
This part is not that hard really, but as with all things AI reliability is one major long term challenge. (source: I helped build Sparrow, DeepMind's RLHF model, that does this by learning to use google under the hood)
201.
by nmca 4y ago
Pasting section 4.2 of this prediction market faq: https://astralcodexten.substack.com/p/prediction-market-faq "What about the risk of insider trading by committing harmful / illegal acts? That is, could Pres
202.
by nmca 4y ago
I'm surprised that nobody links to the best proposed solution to this problem, so I will: https://mason.gmu.edu/~rhanson/dumpceo.html The US could, if it was inclined, legalise prediction markets & investors c
203.
by nmca 4y ago
A huge proportion of crypto projects by count and an embarrassing proportion by market cap are scams. Nonetheless, I really hope someone cracks the remaining crypto issues (scalability, identity, ux, anonymity, governance protocols) well en
204.
by nmca 4y ago
I'm broadly pro harm reduction & happy about psych legalisarion, but the details of the Nutt studies are, aha, pretty Nutty. The most cited one just involved taking some doctors and asking them to yolo out some qualitative harm est
205.
by nmca 4y ago
Isn't this just a different thing? Bouldering builds peak strength , not really intensity in the sense it's meant here? Max hangs do not translate to watts you can do for a minute on a bike, right?
206.
by nmca 4y ago
The best option is make the tests truly fast. The less good option is interlacing two independently testable but closely related tasks. The mind killer is anything else.
207.
by nmca 4y ago
Does anyone have good analysis of what one should expect widely available UBI to do to prices?
208.
by nmca 4y ago
Some folks seem to think the benefit of pomos is the focus time. When I used them, it was the breaks. Stopping every twenty five minutes and asking "is this the right thing to be doing now" was invaluable.
209.
by nmca 4y ago
It has been a wildly surprising fact over the last few years that attenion basically is all you need, in a sense much stronger than the one intended by the authors.
210.
by nmca 4y ago
Do a PhD at ETH Zurich

More ›