Artificial Intelligence / Digital life / Skynet megathread

[dada]

Yes, these are two different things. The AI tech of automated weapons systems, and the artificial super-intelligence indistinguishable from an alien super-intelligence.

AI tech will always show bias, as you point out. And miracles will be marketing. Miracles dreamed of, and never dreamed possible.

[DrEvil]

Computer scientists: We wouldn't be able to control super intelligent machines

Draft Congress report claims AI will make fewer mistakes than humans and lead to reduced casualties

Daft Congress?

They're not wrong. After the super intelligent AI kills everybody casualties will be reduced to zero.

[thrulookingglass]

The (vile) forces of warfare have sought ancient religious weaponry, the ark of the covenant, spear of destiny, had seances with alien/questionable entities in order to obtain high technology, experimented in mind control, biological weapons, chemical weapons, even considered poisoning entire food supplies. They operate in the dark with impossible to trace budgets and have for decades. What makes you think they haven't created this synthetic intelligence already?

This is why you don't allow the war machine to reek havoc upon the world. Nothing is more worthless than the pursuit of war.

[DrEvil]

What makes you think they haven't created this synthetic intelligence already?

They have. They're called corporations and they run on CapitalismOS.

[Grizzly]

Tera-mining Your Micro-infringements (Algorithmic Censorship)


This could just as well fit under the Re: Scary / Trying Times thread....
http://www.rigorousintuition.ca/board2/ ... 34&t=41996

Welcome back for another episode of Nick’s Non-fiction with your host Nick Muniz!
Ian Ayers bestseller, Super Crunchers, is a deep dive into data and the devout power quantum computing will have over our future. Did you know more motor accidents are due to drowsy driving than drunk driving? Insurance companies plan on using micro sensors to determine an increases on your premium based on how many lanes you drift across and dreams you drifted into. Super computers are often super douches to those paying most for its service. Airlines are phasing in smart rebooking to upgrade customers who’re most at risk for continued business rather than existing club members. The book arises a Phillip K. Dick contemplation of technology, ‘Is our sole evolutionary purpose to produce technology, passing the baton to the AI, or do we make machines to serve man’? Ayer crosses the pc line into the forbidden science of IQ deviation and the inconsistencies between the sexes. Ex-president of Harvard, Larry Summers was impeached for recognizing this data going to show even at the apex of academia your future is in the hands of the Super Crunchers.
Subscribe, Share, Mobile links & Time-stamps below!

0:00:45​ Introduction
0:03:25​ About the Author

0:06:25​ Ch1: Who’s Doing Your Thinking for You?
0:13:45​ Ch2: Creating Your Own Data
0:21:25​ Ch3: Government By Chance
0:27:20​ Ch4: Evidence Based Medicine
0:33:55​ Ch5: Experts vs. Equations
0:38:30​ Ch6: Why Not?
0:46:00​ Ch7: Are We Having Fun Yet?
0:51:55​ Ch8: Future of Intuition

0:59:25​ Next Time & Goodbye!

[DrEvil]

Yesterday I downloaded and installed "Stable Diffusion", which is one of the AI image generation algorithms that have been doing the rounds. Being able to run it locally and generating new images in a matter of seconds has been an eye-opening and frankly mind-blowing experience. I haven't had my jaw drop like this over a piece of software since the first time I played Doom as a kid.

Have a look here for some of the stuff people are making with this:
https://www.reddit.com/r/StableDiffusion/

For anyone interested, Polyfjord has a pretty straightforward tutorial on how to install a version with a user interface, and how to upscale the results with another AI program (you need a decent graphics card though).
https://www.youtube.com/watch?v=0zQyCihHjQU

I also installed another version with the img2img model included, which lets you use an image as the input instead of text, so you can iterate and refine images (mind further blown). The installation is a bit more involved, with Python and command lines and stuff, but here's a guide (this one also has the text input version, so if you install this I'm pretty sure you don't need the one from Polyfjord's guide):
https://github.com/hlky/stable-diffusio ... stallation

If you do try this, make sure to have a few hours on hand, because it's addictive as hell. Imagine "This person does not exist", but you can tell it to make anything, like "Angry coked up badger", which I heartily recommend you do.

Also, caveat: I just started messing around with this, so I have no idea if the versions I linked are the best ones out there. Things are moving incredibly fast, with updates and improvements popping up daily.

[cptmarginal]

Stable Diffusion is just what was needed. I was thinking that imagen-pytorch trained on LAION was going to be the first actually open project to rival that of "OpenAI" (a misnomer) but stable-diffusion beat them to the punch. imagen-pytorch is eventually going to support the generation of up to 25 minutes of video, though!

Got to pull myself away from making disturbing images and think of some constructive uses



BTW anyone can play with higher-res outputs and step settings right here, without the need for a nice GPU:

https://beta.dreamstudio.ai/dream

If you run out of credits just login with a different Google or Discord account (or make a burner Discord account)

[DrEvil]

DEAR GOD!

Oh, well. Wasn't planning on sleeping tonight anyway.

Not sure how practical this is right now, other than making nightmare fuel like above, or as a source of inspiration, but that will probably change fast. There's new versions coming out all the time, all of them improving on one thing or another. The 1.5 version of Stable Diffusion is supposedly much better at giving people the correct number of limbs for instance.

Just spitballing, but right now it takes around ten-ish seconds to create an image on a high-end GPU like the 3090. Applying Moore's Law you end up at about ten frames per second in ten years. Throw in improvements to the algorithms, which often outpace Moore's Law, and we're probably looking at real-time video in the near future. Add another few years to work out the nightmare kinks and you can tell your computer to show you the latest Avengers movie, but with everyone naked, or to replace the villain with Donald Trump, or to make up a completely new movie.

[Harvey]

Can you tell me what I do with the files in the zip file on that Git Hub page, please? I can't find any instructions.

[Harvey]

Wow...

[Grizzly]

^^^^
Surreal good job!

-----

[cptmarginal]

[Harvey]