My bet is that we’re at least one breakthrough away from actual artificial intelligence. Not helping is of course the fact that we don’t know what intelligence actually is.
Also, according to Eliezer Yudkowsky, the way we’ll notice a superintelligence moving against us is that all humans will die at once. Of course, while AI with human level intellect is probably possible, on account of humans having human level intellect, superintelligence being possible is still a stretch.
Better, actually. This feeds the crawler a potentially infinite amount of nonsense data. If not caught, this will fill up the whatever storage medium is used. Since the data is generated using Markov-chains, any LLM trained on it will learn to disregard context that goes farther back than one word, which would be disastrous for the quality of any output the LLM produces.
Technically, it would be possible for a single page using iocaine to completely ruin an LLM. With nightshade you’d have to poison quite a number of images. On the other hand, Iocaine text can be easily detected by a human, while nightshade is designed to not be noticeable by humans.
If you use natural text to train model A, and then use model A’s output, a, to train model B, then model B’s output will be less good than model A’s output. The quality degenerates with each generation, but the it happens over generations of models. So, random data is worse than AI slop, because random data is already of the lowest possible quality for AI training.
Just insert motherducker random profanity in your text.
What do you mean by “might need to move on from binary computers to emulate it efficiently”?
Current neural networks do really fancy statistics. To make the model better, you need to make the statistics more precise. Leading to marginal improvements of accuracy requiring exponentially growing marginal amounts of training data. This leads to exponentially decaying marginal utility coupled with exponentially growing marginal expense. Which quickly becomes unsustainable. Edit: On the plus side, this likely means you won’t have to give up much utility when the market adjusts.
GRUB works so well that the average Linux user likely never has to think about its inner workings. Even installing Linux has become extremely easy, like unless you use something like Arch Linux. So, its actually quite likely that somebody who writes a program that runs bash commands would not now how to maintain GRUB.
Neural networks have existed since the 70s.
Is it supposed to be a stuffed animal?
There’s several flavors of Arch?
I tried arch once. Eventually, my computer just showed a black screen on booting. I managed to fix it by resetting my bios. That was the end of that attempt at using arch. Still want to try again, though.
Which, ironically, makes it a toy.
I do want to find the most relevant topics at the top of a community, though.
If you post “I need that on a tshirt” on a twitter post where an artist showcases their work, the image will be automatically plagiarized by a bot and put on a tshirt. The fact that this is profitable shows that intellectual property laws fail to protect vast majority of artists. (https://www.reddit.com/r/tumblr/comments/172vq82/i_want_this_on_a_tshirt/ I apologize for the reddit link).
One argument I’ve heard from the more ancap types is that these things are okay because intellectual property is a scam the stifles innovation. Let’s take that at face value.
IP abolitionism isn’t exclusive to Ancaps. And you don’t even address it again in the remainder of your post, instead attacking a concept ancaps typically use to justify racism.
Why is it wrong to take an image and put a filter on it, assuming you’re honest about what you’re doing?
Yes, the idea is good, I just don’t trust AI to do a good job.
This is more manpig, I don’t see the bear.
I want to read “And Ned Fas Viel”.
Don’t let the door hit you on the way out.