Y
HN Search
Hacker News Search
new
|
comments
|
top
|
jobs
AlanYx
searching PlanetScale…
1.
▲
2.
▲
3.
▲
4.
▲
5.
▲
6.
▲
23 ms
·
301.
▲
by
AlanYx
3y ago
>in direct sunlight on a hot California day you can see a world of difference between the contrast and glare of a high school graphing calculator and a Kindle. Sharp Memory LCDs are reflective, not transreflective. They're easier to
302.
▲
by
AlanYx
3y ago
It's interesting that there are such different perspectives on whether to knead sourdough. The book linked to here recommends it, while other authors like Forkish, Prueitt and Robertson rely more on time and a small amount of "fol
303.
▲
by
AlanYx
3y ago
This is truly phenomenal work! Have you experimented with any traditional additives, such as adding a small amount of citric acid to increase diacetyl production during fermentation?
304.
▲
by
AlanYx
3y ago
For someone not educated as a scientist, he has a remarkably solid history of academic publications. Pretty inspiring actually.
305.
▲
by
AlanYx
3y ago
The CRTC did order CHOI-FM to be shut down, and they lost the lawsuit and all appeals, so they would have been shut down. But at the last minute they were bought by another channel with a different CRTC licence, which saved them from the sh
306.
▲
by
AlanYx
3y ago
This latest proposed ban is significantly more serious than the CHOI-FM dispute, in the sense that they're now considering banning an entire channel because of a complaint that one particular program on that channel was allegedly in vi
307.
▲
by
AlanYx
3y ago
This article focuses on the impact of the EU AI Act on open source AI work and research, but the act itself applies broadly to all development of AI, public and private.
308.
▲
by
AlanYx
3y ago
Some of their current generation plastic fasteners are ingenious and probably the product of substantial mechanical design expertise. Their Dundra table (the Ikea "Lego table") is a good example... I thought there was no way that
309.
▲
by
AlanYx
3y ago
During his lifetime, that's true. However, there was a period in the first part of the 1800s when his works flirted with popular acclaim, especially after Mendelssohn revived the St. Matthew Passion in 1829.
310.
▲
by
AlanYx
4y ago
The book itself recommends that most children start the program in the book at age 4, also noting that some advanced students may be ready at age 3.5.
311.
▲
by
AlanYx
4y ago
In North America, Rousseau has given way to Paulo Freire and successors, but it's the same thing for the most part.
312.
▲
by
AlanYx
4y ago
I just tried your example in GPT-4 and got the correct response: >The man should tell the blind man to "pull" the door to open it. The reason is that the man sees the word "PULL" in mirror writing, which means it is w
313.
▲
by
AlanYx
4y ago
Neal Stephenson also recently reconfirmed that he still uses emacs for writing. I wonder if he's using it on a vintage machine or something modern?
314.
▲
by
AlanYx
4y ago
>no machine translation system currently in existence could translate the 逆転裁判 games to Ace Attorney games Maybe it's already in the training set, but GPT-4 does give that exact translation. I've found that GPT-4 is exceptional
315.
▲
by
AlanYx
4y ago
>Not much of a problem if you have an accounting department next to legal department on the same floor of your corporate building. It's demonstrative to read the official EU impact assessments for a lot of their recent legislation.
316.
▲
by
AlanYx
4y ago
The new EU AI Act also has this problem, in that it imposes liability for developing components that may at some future point be misused by others. The source of the problem is a particular approach to legislation that has become popular in
317.
▲
by
AlanYx
4y ago
Part of the problem is that there are multiple slow "art sequences" in the film that seem to have no clear purpose and run counter to Tarkovsky's usual dislike of intentional symbolism in film. Some seem like they might have
318.
▲
by
AlanYx
4y ago
>But this is a domestic regulator, making a domestic finding. It's a decision by a domestic DPA based on the GDPR. In practice, it will have an impact across the EU, because any company intending to operate uniformly across the EU h
319.
▲
by
AlanYx
4y ago
Just judging from the volume of papers, it seems to me that there are more academics writing papers on "AI safety" and "AI ethics" than there are academics publishing research papers on actual AI. It's become one of
320.
▲
by
AlanYx
4y ago
The sun's spectrum is relatively smooth, whereas there are some grow lights that have a narrow, intense spike of UV designed to match chlorophyll B's absorption. Not all grow lights have this spectrum, but it's worth being ca
321.
▲
by
AlanYx
4y ago
Tomatoes were also viewed with suspicion initially in Italy and used only ornamentally. It wasn't until the 18th century that tomatoes really became accepted in Italian cuisine.
322.
▲
by
AlanYx
4y ago
It's likely Part 3 of bill C-27 (the "Act respecting artificial intelligence systems and data used in artificial intelligence systems"). Google isn't going to launch Bard in Canada if they'll have to pull Bard once
323.
▲
by
AlanYx
4y ago
Version 4 is definitely better in discussing philosophy in general. 3.5 was able to summarize philosophical material, but once you started to engage with ideas critically, it would tend to get into largely nonsensical moralism about needing
324.
▲
by
AlanYx
4y ago
One thing that LLMs have made me realize is just how ungrounded a lot of mainstream academic philosophy was in the 70s and 80s. For example, so much of Derrida's work centered around the impossibility of truly communicating shared mean
325.
▲
by
AlanYx
4y ago
I think that's a very useful distinction. What becomes somewhat obvious after spending any length of time with ChatGPT4 is that it is limited to a particular ethical frame that is unfortunately quite rigid and by no means universal, an
326.
▲
by
AlanYx
4y ago
I've spent a lot of time comparing ChatGPT4 to DeepL Pro on English to French translation. ChatGPT4 is quite good, but not as consistently good as DeepL. But the interesting thing IMHO is the nature of the mistakes ChatGPT makes... oft
327.
▲
by
AlanYx
4y ago
Is there a good paper you can point to that describes/explores this dichotomy? (i.e., AI safety has two branches, AI ethics and AI alignment) I'd like to understand this better. I've been to a couple of "AI ethics"
328.
▲
by
AlanYx
4y ago
What's most interesting about this article IMHO is that these results were found with "long-term cannabis users" defined as those who use "typically 1-4 times weekly". That seems like relatively casual use.
329.
▲
by
AlanYx
4y ago
Lowest quality is probably the same as Apple Music's standard quality: 256kbps AAC (lower on Apple Watch).
330.
▲
by
AlanYx
4y ago
CBT isn't really appropriate for a big chunk of people with serious anger management issues... that was one of the motivators for the development of DBT.
More ›