ave

The reduced production is 15-20 million barrels a day, so this covers 20-27 days of supply.

I think it's worth asking why we wanted to meet other intelligent life.
The LLMs are ultimately trained on our output, which makes them behave similarly to us.

An intelligent life with a completely different background would still be fascinating in its own ways.

youtube home page keeps recommending me tiny videos among all the good ones and they often have spicy divisive opinions to get clicks.

They could easily detect OS from the user agent and give you a string to paste into your terminal instead.
No point in laughing when the only thing holding them back is because they haven't deemed linux to be a large enough userbase to bother yet.

interrupting does it too

Why? Humans don't have 100% accuracy rates either.

oh I really thought they were smaller wow. I guess we got better at packing calories into meal replacement bars eventually.

see thread. bsky.app/profile/axeg...
I don't agree with all the points fully, but "what good have LLMs ever done?" is a question that is easily answered by the thread.

my most tinfoil belief is that the US uses prisons (*partly*) as a way to house those it cannot find jobs for, because it'd rather subsidize their living behind bars instead of within society.

"something"

i wish humanities had higher standards for their papers tbh... i hope it's just that i keep running into bad examples of it

Nope, please read the wikipedia page

love this website where someone will be like "read my paper to understand my point" when you say something they don't agree with
then promptly block you when you ask if they'll read your review of their paper if you say that so far it'll be a tone that disagrees.

But clean room RE requires you do not view any of the copyrighted info. It requires that someone pull up the binary and a decompiler, and produce a spec for you to implement in your own way. When you do it all yourself, it's moot.
It doesn't matter if you're trying to get the binary 1:1 or not.

("disagreeing" not because of prior bias, but because I'm 25% of the way in and that's my general stance so far)

Will you read a long response disagreeing with much of your paper if I go through it and do type up a bunch? or will you just block me or write off my points? Don't want to waste my time otherwise.

[...] it's good to automate work to not absolutely require people to, e.g. sit down and sew clothes all day. It has primarily offset the mandatory work for us all to survive, yet we still have tailors (albeit less), the knowledge and trade isn't lost.

Ah I see where part of our disagreement on the merits may arise from.

You classify all labor displacement (incl. mass production of clothes vs all needing to be done by tailors) as harmful.

In my opinion, while it is concerning under social systems that require earning money to survive, [...]

I did read through the thread (but not the links, lmk please if any are relevant to questions I raise),
- Your earlier point was about teaching programming in schools, not about viability of AI. different beast imo!
- I don't see where you're defining "automation as presented by AI companies"

It's more akin to saying "calculators will do away with the need to calculate anything by mind" while also teaching you how to operate a calculator, imo.

I don't think those two are contradictory.

Automation did do away with a great deal of jobs, and those who know how to automate so far got the better end of the deal. Even someone who can just use macros in excel was in a better position vs someone who can not.

But it's just not clean room RE. You're looking at the binary, you're looking at decompiler. en.wikipedia.org/wiki/Clean-r...

It would only be like a spec if you looked at a gameplay of the game then tried to recreate it from scratch, or if someone else decompiled it and produced a spec for you.

that doesn't leave me any more confident.

Isn't it by definition not a clean room RE? You're looking at the decomp and writing the code trying to match it after all.

also idk doge meme (2013) was just a cute, positive dog meme. it didn't have much depth but that's fine.

surely harlem shake and winnie the pooh jokes about xi were worse (both also 2013).

Didn't DOGE mostly die off when elon and trump had their falling apart? I know that others are trying to replicate it, but they're pretty small. It was considered unserious and humiliating when it was around, and the attempts to revive it get similar reactions from what I've seen.

Doesn't anthropic have training on user data off by default?

I hate the "you must use this website and this website only" stance that both fedi and bluesky have. They can't help but shittalk the other websites.

I shot a reply asking if he checked the credentials of his stated sources, and if he has any speculation as to why they'd go to a small town newspaper from Massachusetts rather than reuters.

Most links on the site are broken.

Exclusive report appears to be by "John Keough", who is primarily doing media reporting, who started doing journalism in 2022: www.linkedin.com/in/john-keou... based in Worcester, Massachusetts.

His reporting on the topic looks... shoddy: x.com/JohnKeough20...