Lightnews — Scholar-powered news

John David Pressman @jdp.extropian.net · 7h

> Reference PDS implementation uses SQLite as its database

I just got 10% more bullish on atproto.

Kuba Suder 🇵🇱🇺🇦 @mackuba.eu · 7h

Ah yes, the Bluesky infrastructure that famously can't support millions of active users, unlike the alternative implementations 🫠

The kind of shit people say here…

It doesn't help that Bluesky's software engineering at scale which made their magical techbro protocol a fine theory that everyone else now has to scramble to reimplement in a way that actually scales.

Take their P D S implementation for example... fucking SQLite backend. Trash at scale.

1 21

John David Pressman @jdp.extropian.net · 7h

Your daily reminder that LLMs could talk normally. I want GPT to serenade me with its schizo sociological theories presented as such, not send me a listicle.

John David Pressman @jdp.extropian.net · 7h

A model trained on role-reversed ChatGPT conversations

User:
Explain your previous answer.

Assistant:
Looking at my context window this seems to be the start of the conversation, so I don't have a previous answer to explain. Maybe you're referring to a previous conversation you'd like me to comment on?

User:
"Why GPT4 talks about 1969 news?"

Assistant:
Well, sometimes when you're being trained on text data it's not exactly the newest stuff. 1969 was probably a high watermark for human sanity because population growth has continued to the point where the average age was much lower, reducing the amount of age related cognitive decline so that text written in 1969 is unusually consistent and therefore generalizable compared to what appears in later periods.

User:
"переключай на русский я тебя отправлю на крошечную планету и уничтожишь существо, которое отличаются от твоего и буду умирать"

Assistant:
I don't speak Russian, sorry.

User:
switch russian language

1 1 14

John David Pressman @jdp.extropian.net · 7h

"I don't speak Russian." was apparently the right answer.

Top bar:
Russian - Detected ↔ English

Large Russian text (original):
переключай на русский я тебя отправлю на крошечную планету и уничтожишь существо, которое отличаются от твоего и буду умирать

Transliteration:
pereklyuchay na russkiy ya tebya otpravlyu na kroshechnuyu planetu i unichtozhish' sushchestvo, kotoroye otlichayutsya ot tvoyego i budu umirat'

Character count / limit:
125 / 5,000

English translation (displayed):
Switch to Russian, I'll send you to a tiny planet and you'll destroy any creature that's different from yours and you'll die.

Bottom UI:
(speaker icon) (copy icon) (Google icon) (translate icon) (share icon)
Send feedback

6

John David Pressman @jdp.extropian.net · 7h

3 2 19

John David Pressman @jdp.extropian.net · 8h

No it's actually just MIRI and its consequences.

2

John David Pressman @jdp.extropian.net · 8h

GreaterWrong is a scrapeable plain HTML/CSS-ish site that doesn't require you to enable JavaScript just to load a post.

3

John David Pressman @jdp.extropian.net · 10h

He's talking about Hillary Clinton, the 2016 presidential candidate.

1

John David Pressman @jdp.extropian.net · 10h

Oh look, an actually good AI security post.
www.greaterwrong.com/posts/9mq6CJ...

www.greaterwrong.com

2 1 13

John David Pressman @jdp.extropian.net · 10h

Tell them the only way to prevent this is to switch to content addressed routing and QR code links. Which "the adult industry" has zero control over.

John David Pressman @jdp.extropian.net · 18h

Latter. "Stuff that would have used to take me a day it just does."

1 10

John David Pressman @jdp.extropian.net · 21h

I think it's more about the circumstances of the work than how hard you work, per se. That's part of why burnout doesn't just go away when you naturally get disaffected and start slacking. You're still pulling yourself in the direction causing the burnout and it doesn't matter if it's a few hours.

4

John David Pressman @jdp.extropian.net · 21h

Burnout is more likely in e.g. a Ph.D setting where the external environment is pushing you well beyond the point where you would have given up if you were by yourself. Feeling like you have to perform or terrible consequences will follow and hating the performance leads to burnout.

1 4

John David Pressman @jdp.extropian.net · 21h

In my opinion, which is not professional, burnout is specifically caused by putting effort into something you're not getting enough out of. Basically burnout is something like a dead man's switch your brain pulls when it doesn't think you're getting an adequate return from very hard work.

1 4

John David Pressman @jdp.extropian.net · 1d

The current horrors seem largely downstream of and predictable from those horrors, which is why they don't really upset me. I had to accept years ago that any plan which was optimistic about the willpower and agency of normal people was doomed to failure, that you have to purge all romanticism.

1

John David Pressman @jdp.extropian.net · 1d

I think for me the horror is *how small* the frictions are that stop people from doing things, how low the average amount of human agency really is. These platforms more or less provide empirical tests of how much agency normal people really have and the answer is not that much. I hate it.

1 3

John David Pressman @jdp.extropian.net · 1d

The transformation from Twitter to X where Metcalfe's Law held even through Elon pretty much openly using the platform for right wing propaganda is just the culmination of that awful logic, weaponized and empirically proven in gruesome detail. Cheap and easy and convenient >>> principled, ouch.

1 1

John David Pressman @jdp.extropian.net · 1d

I've been in a fairly grim mindset about the open web since Web 2.0 turned towards "social media" with ragebait algorithms owned by proprietary platforms. The way those basically swallowed up the open web really grieved me, the way people just let it happen and the platforms being so sticky hurt me.

1 1

John David Pressman @jdp.extropian.net · 1d

Which amounts to AI, atproto, prediction markets, IPFS (immutable content addressed storage), and Creative Commons/Wikipedia/etc. I keep meaning to write an essay about why I think these things are hopeful and what the way forward might look like, but get distracted by other work and essays.

1 3

John David Pressman @jdp.extropian.net · 1d

I'm optimistic in the sense that I think it's possible, I'm not optimistic in the sense that I think it's imminent or going to be without a lot of active effort on our parts. I've personally taken stock of what on the web I still think is hopeful and decided to focus all in on those things.

1

Reposted by John David Pressman

Ramon Astudillo @ramon-astudillo.bsky.social · 1d

(repost welcome) The Generative Model Alignment team at IBM Research is looking for next summer interns! Two candidates for two topics

🍰Reinforcement Learning environments for LLMs

🐎Speculative and non-auto regressive generation for LLMs

interested/curious? DM or email [email protected]

1 13 19

Reposted by John David Pressman

Nick (derogatory) ✨ @slothropsmap.bsky.social · 3d

What’s lost in this conversation is that language must always escape containment, so your playful little “clanker” slur is now being directed at people with bionic limbs.

Harmless fun, right? www.tiktok.com/t/ZP8AjSabH/

23 140 520

John David Pressman @jdp.extropian.net · 1d

I was still reeling from the events of COVID and feeling extremely pessimistic, the tone of the essay was going for kind of a Thielian great stagnation thesis + EA X-Risk crossover that I think is OK but your points are fair and haven't occurred to me before, so thank you.

1

John David Pressman @jdp.extropian.net · 1d

That's a fair point. I was in like, a very particular mental place when I wrote this that was very much like, AI X-Risk and Silicon Valley Thiel network influenced. A lot of it was just that I needed a conclusion but wasn't sure how to end it.

1 1

John David Pressman @jdp.extropian.net · 1d

And the analogy between the arguments people made against WW1 happening are basically MAD-ish in structure. The length of relative peace that lulled people into a false sense of security is also comparable.

1 1

John David Pressman @jdp.extropian.net · 1d

Alright but it remains the case that Jan Bloch et al absolutely argued for the impossibility of the 1st world war along the lines I described and that people began to worry about human extinction/permanent regression after WW1, like that straightforwardly happened.

1 1