KolmogorovComplicity
No bio...
User ID: 126
KolmogorovComplicity
2yr ago
It's not like these algorithms are generating inhuman images for their own inhuman purposes and flooding the Internet with them. Every image produced by one of these algorithms is something a human requested, and, if they bother to share it, presumably finds valuable in some way. That's still firmly within "human culture."
KolmogorovComplicity
2yr ago
Red tribe, to the extent much of their jobs include manipulating the physical world directly, may turn out to be relatively robust against AI replacement.
Perhaps, but look at DayDreamer:
The Dreamer algorithm has recently shown great promise for learning from small amounts of interaction by planning within a learned world model, outperforming pure reinforcement learning in video games. Learning a world model to predict the outcomes of potential actions enables planning in imagination, reducing the amount of trial and error needed in the real environment. [...] Dreamer trains a quadruped robot to roll off its back, stand up, and walk from scratch and without resets in only 1 hour. We then push the robot and find that Dreamer adapts within 10 minutes to withstand perturbations or quickly roll over and stand back up. On two different robotic arms, Dreamer learns to pick and place multiple objects directly from camera images and sparse rewards, approaching human performance. On a wheeled robot, Dreamer learns to navigate to a goal position purely from camera images, automatically resolving ambiguity about the robot orientation.
Stable Diffusion and GPT-3 are impressive, but most problems, physical or non-physical, don't have that much training data available. Algorithms are going to need to get more sample-efficient to achieve competence on most non-physical tasks, and as they do they'll be better at learning physical tasks too.
KolmogorovComplicity
2yr ago
Logically, shouldn't we expect powerful absolutist/totalitarian states to dominate, ceteris paribus?
Totalitarian systems tend to suppress innovation, either deliberately because the powerful fear the social changes they might produce, or unintentionally by restricting ideas and freedom of action. For most of history, innovation was sufficiently slow that this wasn't important, or at least took centuries to catch up with a given society, but the industrial revolution rapidly accelerated the process. Today a nation can become economically and militarily uncompetitive in as little as 10 or 20 years. That's fast enough to register on the planning horizons of current leaders.
China has done well for itself economically over the last few decades, but this was mostly catch-up growth, the adoption of already-existing tech. Here, there's no need to have a society that fosters innovation through freedom of action and a free exchange of ideas, since you're merely deploying the products of innovation that took place elsewhere. There's also much less risk of social disruption, as you can look at the social changes that particular technologies created in countries that deployed them earlier, and shape deployment to ameliorate those (see e.g. the Great Firewall). Incrementally improving an existing technology, such as by refining the manufacturing process for an existing product, has similar properties.
China has yet to demonstrate it's capable of fundamental innovation, of being first to invent and deploy a basically new thing. So far, the signs don't look too great — a tech industry crackdown, a cryptocurrency ban, a requirement for government pre-approval of individual App Store apps. It's hard to believe China won't carefully restrict AI capabilities to a few trusted institutions. I could see them objecting even to something as anodyne as Stable Diffusion before too long, given that it will happily generate as many offensive caricatures as you'd like of Xi Jinping.
This is intended to make comment threads more readable, primarily by drawing borders around comments so the nesting structure is more obvious. Also adjusts comment thread whitespace. The last rule limits the bodies of posts and comments to a reasonable width, so lines of text aren't uncomfortably long on large screens. Only tested with the default theme, and better tested on desktop than mobile. Screenshot attached.
Edit: now with proper margins for the 'more comments' buttons that appear for deeply-nested posts.
.comment .comment-collapse-desktop, .comment .comment-collapse-desktop:hover {
border-left: none !important;
background-color: var(--gray-400);
padding-right: 7px;
border-radius: 7px 0 0 0;
}
.comment .comment-collapse-desktop:hover {
background-color: var(--primary-light1);
}
.comment .comment-body {
border: 1px solid var(--gray-400);
border-left: none;
padding: 0;
}
.comment, .comment-section > .comment {
margin: 1rem -1px -1px 0;
padding-left: 0;
border-color: var(--gray-400) !important;
border-width: 5px !important;
border-radius: 5px 0 0 0;
}
.comment .comment {
margin-left: 1rem;
}
.comment-anchor:target, .unread {
background-color: rgba(0, 230, 245, 0.1) !important;
}
.comment-write {
padding: 1rem !important;
}
.more-comments > button {
margin: 1rem !important;
}
#post-text, .comment-text, .comment-write {
max-width: 60rem !important;
}
You can also add this rule if you want to change the font weight and size for post/comment bodies:
#post-text, .comment-text, .comment-write, #post-text p, .comment-text p, .comment-write p {
font-size: 16px;
font-weight: 450;
}
I believe the defaults are 14px and 400.
The best online discussions I've had over the 20+ years I've been having them have almost all been in old phpBB-type forums or (further back) on Usenet, where there were no scoring systems. I don't believe this is a coincidence. Even though rationally people shouldn't care that much about fake Internet points, they do, and there's a tendency to pander to an understood consensus, either by not raising arguments you think will be unpopular in the first place, or by prematurely terminating exchanges where you've discovered the consensus opposes you.
So my preference would be to simply eliminate voting, or, failing that, to hide comment scores from non-moderators, including from comment authors.
More options
Context Copy link