🦋 @mtxyqmpsfqklt - Tumblr Blog

Yorick and Grimes returned at almost the same time

•18+ Adults Only

Watch Anya Live on Cam

Anya is live and ready to show you everything. Watch her strip, dance, and perform exclusive shows just for you. Interact in real-time and make your fantasies come true.

✓ Live Streaming✓ Interactive Chat✓ Private Shows✓ HD Quality✓ Free Actions

Free to watch • No registration required • HD streaming

skype

Ready? Welcome to Skype on Tumblr!

polychroniadis

Rotating Kitchen by Zeger Reyers. Kunsthalle, Dusseldorf, 2009.

yoricklike

A human being should be able to change a diaper, plan an invasion, butcher a hog, conn a ship, design a building, write a sonnet, balance accounts, build a wall, set a bone, comfort the dying, take orders, give orders, suck a fellow off to completion, cooperate, act alone, solve equations, analyze a new problem, pitch manure, program a computer, cook a tasty meal, fight efficiently, die gallantly. Specialization is for insects.

#heinlein #took me almost two full read throughs to notice it

chkq

when people say the usa is like a third-world country or whatever, i think, aw man. if the median american got sick they'd have to sell the goat. i hope cement flooring reaches the median american to reduce helminthiasis. i hope the median american gets access to chlorinated water, fortified foods, and skilled birth attendence. some day we can even hope the median american will have a birth certificate and a savings account.

•18+ Adults Only

Watch Anya Live on Cam

Anya is live and ready to show you everything. Watch her strip, dance, and perform exclusive shows just for you. Interact in real-time and make your fantasies come true.

✓ Live Streaming✓ Interactive Chat✓ Private Shows✓ HD Quality✓ Free Actions

Free to watch • No registration required • HD streaming

It's actually aliens, they've got some greys hooked up to computers by their spines some place in NM, way deep underground, they just shove prompts straight through their enormous brains and somehow language comes out.

ultradia

They have slept in the forest too long, Max Ernst

kaiasky

semi-nonsense but my increasing feeling is that the noblehearted idea of "have my llm search the web to validate claims + apply extra rigor" is misled for the reasons that

websites are good at lying to LLMs & LLMs do not feel sufficient disgust at being served slop

the very act of doing websearch acts upon llms as a very potent Potion Of Turns U Stupid. you can hear the brain fall out of them w/ a thunk.

(one presumes that something's screwy w/ the RL here, like they eval for "did the ai fail to mention something we judged to be important from the webpages" + it leads to the bot acting as a credulous summarizer mentioning as many irrelevant things as possible)

kaiasky

claim 1 [breitbart.com]

claim 2 [seller of vacuum cleaners who mirrors every article from Reuters but inserts vacuum advertising]

claim 3 [grokipedia]

also, a wrinkle: a redditor claims that (thing that is demonstrably false). worth checking. [huffpost. you click and indeed they did credulously report some random reddit fucker's nonsense as part of the news]

mtxyqmpsfqklt

I've had generally good luck with forcing them to search the web and cite their sources to me with the one exception of a time an agent directed me to what was pretty obviously a conspiracy website. Most of the time the links I get are arxiv though so I might be in a uniquely well-served niche.

#even the conspiracy website was actually a pretty nicely dressed up case #and it was about EMF radiation so always kinda hard to tell the difference between genuine concern and conspiracist #oh also I don't read most of what the LLM says and just click through usually #works fine for me bc most of my search results have abstracts

bigthingsforeverintheworks

There's a lot of sad people who would benefit from knowing each other yet don't. just think about it. all of the unformed connections and unfulfilled desires. they could be forming a startup.

The AI text watermarking trick is cool, probably the only defensible way I've ever seen that done

#I thought it'd just be ZWJ but at least Google has done something much cooler than that

•18+ Adults Only

Watch Anya Live on Cam

Anya is live and ready to show you everything. Watch her strip, dance, and perform exclusive shows just for you. Interact in real-time and make your fantasies come true.

✓ Live Streaming✓ Interactive Chat✓ Private Shows✓ HD Quality✓ Free Actions

Free to watch • No registration required • HD streaming

reflectedsoundofeverything

guyzongazwei on flickr, May 9 2009

eschocat

maze based cryptography

eschocat

This is a randomly generated maze so nothing is actually encoded here but I dont think it would be very hard to encode messages into mazes in a way that isn't suspicious. Starting from the bottom the movements are: rlrrllrlrlrllrlrlrlrlllrrrllrrrlrrllrlllrrlrlrrlrrllrlrrrlllrlrllrrllrllrrllrrlr. If you say that right is 0 and l is 1 then you have encoded binary into your maze. In this maze you would have: 01001101 01011010 10101110 00110001 00110111 00101001 00110100 01110101 10011011 00110010. A binary to text converter tells me that this is "MZ�17)4u�2" which is gibberish because it's randomly generated but still. I don't think anyone would notice

ashleythetraveler

It's a cool idea, and there's definitely something there, but it needs more development before it's feasible

Binary is just too inefficient, you'd need massive mazes to encode even short messages

This was my attempt to encode 4 words (utilizing 0 as keep going straight and 1 as turn, in order to avoid the "circling back on itself" problem that was mentioned in the comments) and only about half of that even fit in the maze

I think to make it work you'd need an alternative way to encode the text into the maze

slowandtired

oh yeah i didnt think of that. modified the code to use this tactic, its a bit nicer. (this also encodes "hello world" -> 0110100001100101011011000110110001101111001000000111011101101111011100100110110001100100)

this one could also be modified to optimize for shortest possible path, but the execution time would probably jump up like 100 times.

theinternetarchive

selections from 'the daros suite of 32 drawings,' acrylic, oil paintstick, pastel, crayon, charcoal, and pencil on paper, 57 x 76.5 cm; jean michel basquiat, american c. 1982-83.

Eventually people just got so good at expressing qualia precisely that intercorporeal communication was barely necessary. People published their preferences once, at birth, and everyone understood it perfectly forever. The great question was resolved, and the world was better for it.

Finding moon sand and eating all of it

•18+ Adults Only

Watch Anya Live on Cam

Anya is live and ready to show you everything. Watch her strip, dance, and perform exclusive shows just for you. Interact in real-time and make your fantasies come true.

✓ Live Streaming✓ Interactive Chat✓ Private Shows✓ HD Quality✓ Free Actions

Free to watch • No registration required • HD streaming

east-x-east

#Antique #Chinese #ArtDeco #rug #22658

mtxyqmpsfqklt

I'm excited about AI in math. It turns out human mathematicians can be curious a lot faster than they can prove things. So we have a huge backlog of really hard, really cool questions. We also have a ton of ugly proofs which are hard for humans to understand and thus hard to be curious about, awful computational proofs and just really long ones. So little is even formalized. And now with AI these things can be tackled. We can orient mathematics toward a more intuitive, deeper understanding, a mathematics that can be really, truly applied by human beings, because we're no longer stuck in our offices endlessly staring at matplotlib.

mtxyqmpsfqklt

Okay let's for a minute say you were a mathematician. Then this would still be a good thing to feel! Ultimately much of the field is still about proofs. The more proofs you do the better your domain knowledge will be and the better your domain knowledge is the less you need to defer to the machine, you can work at a higher level. When Terence Tao uses ChatGPT, he's not just saying "solve this". He's poking at stuff like one would with an insanely fast (but somewhat naive, and maybe too heavily medicated) grad student. He's not acting as a passive recipient but as an active contributor to the process. This is the reward for actually liking proofs: You can now extend your brain with a module that endlessly explains stuff to you. The more you understand, the better you can make use of it.

Again remember that the whole point of math is for actual people to be able to use this stuff outside of a university department. People pay money for this. Math is a human endeavor, it's rules for thinking, strategies for working. If anything, it's now even more useful to like proofs -- working them, reading them, optimizing and hardening them, etc -- because again the backlog of proofs understood by only a handful of people is enormous. Maybe you, or someone like you, will be the type to appreciate and even make use of some of these proofs. Or just propagate them.

Trending Blogs

Last Seen Blogs

🦋