tmalsburg2

Publication servers like arXiv + overlay journals. I‘d love that.

People talk as if dynamic scoping was objectively a mistake, but the fact that it works well and is really useful in a complex piece of software like Emacs seems to suggest otherwise.

Napster may have started the file sharing revolution. But the exciting part for me was Gnutella and later Bittorrent, peer-to-peer technology in general, and the realization that we could use technology to liberate ourselves. Needless to say, I was young and naïve. That spirit is long dead, and the only remnants are crypto currencies and the community around them, which has tossed all lofty ideals in favor of blind greed.

Japan's initiative is focused on "green open access" which is different from pay-to-publish. I recommend the section titled "Green OA" in the submitted article. Relevant quote:

"Japan’s move to greater access to its research is focusing on 'green OA' — in which authors make the author-accepted, but unfinalized, versions of papers available in the digital repositories, says Seiichi.

Seiichi says that gold OA — in which the final copyedited and polished version of a paper is made freely available on the journal site — is not feasible on a wide scale. That’s because the cost to make every paper free to read would be too high for universities."

Smart, because why would you use billions in tax money to fund research then let publisher parasites take the results hostage.

A single word is insufficient evidence to conclude that an LLM was used. "Delve" may be low frequency in naturalistic text but there are many words in an article and the chance that some of them will be low-frequency is high. I also checked in my bibliography and found that "delve" is actually not super rare in academic papers including those written before LLMs.

So when there is demand, the thing is depleted after ~1 hour.

My wife is still using her 2012 MBP. We maxed out RAM and gave it an SSD in 2016. She uses it for video editing and music production. The thing look like new. Completely ridiculous. Only downside: no OSX updates since I don’t know when.

OP implied that we‘re experiencing the Bertha Benz moment of electric cars. Wasn’t my idea. I just spelled out the possible implication.

Megawatt seems like the wrong unit as it doesn’t measure capacity. Please provide a link.

Large-scale storage technology is still elusive, though.

After Bertha Benz’ famous drive [1], it took multiple decades for cars to really go mainstream. I wonder if we’re looking at a similar timeline.

[1] https://en.m.wikipedia.org/wiki/Bertha_Benz#First_cross-coun...

I‘ve never met a person who has made their whole personality about being anti-AI. Can we please calm down a bit?

Today's climate, where Silicon Valley billionaires occasionally have to endure some mild pushback.

Didn't they also ask for permission from her two days before releasing the voice? Why ask permission if it was not her voice?

Yes, he got incredibly lucky that he found an alternative use for his GPU investment.

His engineering mindset made him blind to the fact the metaverse was a product that nobody wanted or needed. In one of the Fridman interviews, he goes on and on about all the cool technical challenges involved in making the metaverse work. But when Fridman asked him what he likes to do in his spare time, it was all things that you could precisely not do in the metaverse. It was baffling to me that he failed to connect the dots.

Love how everyone is romanticizing his engineering mindset. But have we already forgotten that he was even more passionate about the metaverse which, as far as I can tell, was a 50B failure?

Fully agree. This stuff doesn’t magically become interesting and relevant to HN just because some connection to a tech company exists.

Wouldn’t the best thing have been to retract the book, or at least to published a revised edition?

Do LLM not also think when they encode the prompt? If Karpathy's explanation is accurate, longer prompts should also help even if they don't contain additional information, just by virtue of giving more time to think.

> - Training isn’t done at 4-bits, to date this small size has only been for inference.

Wasn't there a paper from Microsoft two weeks ago or so where they trained on log₂(3) bits?

Edit: https://arxiv.org/pdf/2402.17764.pdf

I learned a lot about this from the book "The design of the Unix operating system" by Maurice J. Bach.¹ It's an old book and many details deviate from actual present-day Linux, but it nonetheless gives a great overview of the key components and ideas.

¹ https://books.google.de/books/about/The_Design_of_the_UNIX_O...

> Not for electricity production.

A quick Google search suggests ~350g CO2 per KWh in Germany vs. ~400g in the US. Looks plausible to me. Germany uses coal but also has a very substantial share of renewables. Another quick search also shows that 60% of electricity in the US is from fossil fuels. The rest is nuclear and renewables in about equal amounts.

And yet CO2 per capita is much lower in Germany than in the US. So Germany must also be doing something right.

I didn’t accuse anyone. I just laid out the possibilities. Reading comprehension.

+1 for restic. I tried various solutions and restic is the best by far. So fast, so reliable.

https://restic.net/

> Alaska Airlines doesn't trust Boeing enough

They don't trust Boeing enough or their customers don't trust Boeing enough and it's a PR measure to calm the passengers' nerves. Could also be a bit of both.

The answer is in the second paragraph.

My understanding (IANAL) so far was that licenses apply to the code and binaries derived from the code. Why do you think that GPL also applies to a from-scratch rewrite? There are cases where a rewrite was done precisely to circumvent license restrictions.