nmca

7 ms·

91.
by nmca 3y ago
The driving is extremely smooth in general --- much smoother than many taxi journeys I've taken in that past. I found myself in a corner case recently, with an illegally parked car blocking the street, we got partially round, blocked b
92.
by nmca 3y ago
Wonderful new stuff. I heard it suggested recently that it would be interesting to build a static analysis system around proving that code was wrong (as opposed to proving that certain classes of bugs were missing). Fuzzing seems like one r
93.
by nmca 3y ago
Cowen has a counterpoint that whilst finance has grown as a proportion of GDP it has remained very close to 2% as a proportion of wealth, which honestly seems pretty reasonable.
94.
by nmca 3y ago
I haven't yet read this in detail, but it seemed to (reasonably) miss Chatterjee correlation, which is also fun.
95.
by nmca 3y ago
gpt4 first go: If you initially set a timer for 45 minutes and there are 5 minutes left, that means 40 minutes have already passed. To make the total timer time 60 minutes, you need to add an additional 20 minutes. This will give you a tota
96.
by nmca 3y ago
Initially I was confused by the business model here, but I eventually figured it out --- improved python tooling is going to accelerate AI so much that they can just rely on the future gratitude of the machine god.
97.
by nmca 3y ago
best by far != perfect
98.
by nmca 3y ago
By far and away the best python plotting library is plotnine, a python clone of ggplot maintained by Hassan Kibirige from Uganda. By the estimation of one esteemed colleague "it obviously pays back the time investment in less than a co
99.
by nmca 3y ago
The point is that paying more isn't gaming the system (and to be super explicit, yes if the US were to choose to allocate 25k annual h1b by ranked wage & thus all 25k were to go to various flavours of computer programmer or financi
100.
by nmca 3y ago
right, but that's approximately because one is less valuable modulo the adjustment period and some noise.
101.
by nmca 3y ago
Some healthy subset of companies just want to hire the best people they can get that will accept their offer, and sponsoring an h1b is a small part of the overall costs.
102.
by nmca 3y ago
I eagerly await ruff-but-for-type-checking. Python type checkers are all excruciatingly slow at the moment.
103.
by nmca 3y ago
requiring events to kill anyone as evidence of danger of death is a foolish standard, obviously. People are "bad at statistics" in the sense that the very real evidence coming from, e.g. 3 mile, is hard to bring into statistical m
104.
by nmca 3y ago
This is a noble take for programming languages too, where point-free style in e.g. Haskell is a pain in the neck, and named arguments are a blessing.
105.
by nmca 3y ago
By 2030, if not sooner, the idea of humans writing software will be very quaint.
106.
by nmca 3y ago
idk about hierarchical planning just ask an LLM it'll probably work https://say-can.github.io/
107.
by nmca 3y ago
It's a lot better than last year, and techniques seem very general.
108.
by nmca 3y ago
It is pretty funny that this makes use of AI. Even the haters accept the mundane utility.
109.
by nmca 3y ago
> people are not naturally cooperative Not perfectly cooperative, sure, but shockingly cooperative. Just recently I was sat in a commercial airline for 9 hours and struck by how much harder it would be to transport several hundred consci
110.
by nmca 3y ago
You're missing the point, all the things are very bad but have very different relative risks.
111.
by nmca 3y ago
Lazarus Lake on Conversations with Tyler: https://open.spotify.com/episode/75PyLxu7sZa7GkZtOUI3lJ?si=E... (Lazarus Lake organizes ultramarathons. He's a legend. You don't need to be a runner to enjoy this) A
112.
by nmca 3y ago
From the article, and relevant here: I’m worried that when people hear ‘5 OOMs off’, how they register it is, “Oh we have 5x less data than we need - we just need a couple of 2x improvements in data efficiency, and we’re golden”. After all,
113.
by nmca 3y ago
I don't think your comment is true unless you also include a linearity assumption that almost certainly doesn't hold, see e.g. prospect theory.
114.
by nmca 3y ago
https://ourworldindata.org/life-expectancy
115.
by nmca 3y ago
Some hedging possible by investing in AI as opposed to the whole market.
116.
by nmca 3y ago
If you enjoy darker/harder sci-fi "Axiomatic" is a great collection by Greg Egan.
117.
by nmca 3y ago
That was six weeks ago aha, the models work now.
118.
by nmca 3y ago
You're confusing "that one photo of three ceos at the whitehouse" with "all government solicitation of expert opinion on AI". It's correct that the meeting you're thinking of was industry-heavy, but the pr
119.
by nmca 3y ago
None of the world-leading scientists that co-authored this paper: https://managing-ai-risks.com/ Work for big labs? How is that you think the proposed conspiracy works exactly? They're paying off Yoshua Bengio? (discla
120.
by nmca 3y ago
if the position was high-status enough they might be able to hire someone post-economic

More ›