this post was submitted on 18 Feb 2024
173 points (100.0% liked)
Technology
37712 readers
157 users here now
A nice place to discuss rumors, happenings, innovations, and challenges in the technology sphere. We also welcome discussions on the intersections of technology and society. If it’s technological news or discussion of technology, it probably belongs here.
Remember the overriding ethos on Beehaw: Be(e) Nice. Each user you encounter here is a person, and should be treated with kindness (even if they’re wrong, or use a Linux distro you don’t like). Personal attacks will not be tolerated.
Subcommunities on Beehaw:
This community's icon was made by Aaron Schneider, under the CC-BY-NC-SA 4.0 license.
founded 2 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
As I always write, trying to restrict AI training on the ground of copyright will only backfire. The sad truth is that malicious parties (dictatorships) will get more training materials because they won't abide by rules. The end result is, dictators would outperform democracies in terms of future generation AIs, if we treat AI training like human reading.
You know what?
I'm fine with that hypothetical risk.
"The bad guys will do it anyway so we need to do it, too" is the worst kind of fatalism. That kind of logic can be used to justify any number of heinous acts, and I refuse to live in a world where the worst of us are allowed to drag down the rest of us.
Yeah, I mean bad guys are going to commit murder too, doesn't mean it shouldn't be illegal.
The consequence of falling behind is gravely different from most heinous acts. It can impact the military, elections, espionage, or whatever.
Really? I'm supposed to believe AI is somehow more existentially risky than, say, chemical or biological weapons, or human cloning and genetic engineering (all of which are banned or heavily regulated in developed nations)? Please.
I understand the AI hype artists have done a masterful job convincing everyone that their tech is so insanely powerful (and thus incredibly valuable to prospective investors) that it'll wipe out humanity, but let's try to be realistic.
But you know, let's take your premise as a given. Even despite that risk, I refuse to let an unknowable hypothetical be used to hold our better natures hostage. The examples are countless of governments and corporations using vague threats as a way to get us to accept bad deals at the barrel of a virtual gun. Sorry, I will not play along.
If you don't see how even the most basic of AI images, videos, deepfakes, etc. can manipulate the public, the electorate, popular opinion, or even sow just enough doubt as a cause a problem, then I don't know what to tell you.
People are already dying because of deepfakes and fake AI porn. We know that most people who see some headline on Facebook will never click farther to read it, and will just accept the headline and/or the synopsis as fact. They will accept something a 1000x re-shared image says, without sources or verification. The fact that a picture or vid might have a person with 8 fingers on one hand in the background isn't going to prevent them from taking in the message. And we've all literally seen people around the web say , explicitly, something to the effect of "I don't care if the story is true or not, it's a real issue we need to consider" when we know for a fact that it is not.
Yes, mis- and dis-information are far more of an existential thread than chem or bio weapons, and we know this because we are already seeing the consequences of it. If you refuse to see that, then you are lost.
You don't need AI for any of that. Determined state actors have been fabricating information and propagandizing the public, mechanical Turk style, for a long long time now. When you can recruit thousands of people as cheap labour to make shit up online, you don't need an LLM.
So no, I don't believe AI represents a new or unique risk at the hands of state actors, and therefore no, I'm not so worried about these technologies landing in the hands of adversaries that I think we should abandon our values or beliefs Just In Case. We've had enough of that already, thank you very much.
And that's ignoring the fact that an adversarial state actor having access to advanced LLMs isn't somehow negated or offset by us having them, too. There's no MAD for generative AI.
What beliefs and values would we be abandoning by fighting back against tech that is literally costing people their literal lives?
Hah I... think we're on the same side?
The original comment was justifying unregulated and unmitigated research into AI on the premise that it's so dangerous that we can't allow adversaries to have the tech unless we have it too.
My claim is AI is not so existentially risky that holding back its development in our part of the world will somehow put us at risk if an adversarial nation charges ahead.
So no, it's not harmless, but it's also not "shit this is basically like nukes" harmful either. It's just the usual, shitty SV kind of harmful: it will eliminate jobs, increase wealth inequality, destroy the livelihoods of artists, and make the internet a generally worse place to be. And it's more important for us to mitigate those harms, now, than to worry about some future nation state threat that I don't believe actually exists.
(It'll also have lots of positive impact as well, but that's not what we're talking about here)
Ah gotcha. I must have misunderstood the flow there. Yeah, definitely seems like we're mostly on the same side
But, if we make training ai without copyright illegal, it will hamper open source models, while not affecting closed source ones , because they could just buy it off of big social media conglomerates
Alrighty then. If corps want to train their AI on all the content they can scrape without worrying about copyright, then they can't complain when I torrent their shit without worrying about copyright too! Deal? Somehow I don't see them taking that deal.
Training new models is already the domain of large actors only, simply due to the GPU requirements, which serve as a massive moat. That ship has sailed. There isn't a single open source model, today, that wasn't trained by a corporate entity first, and then only fined tuned by the community later.
"Bad guys are going to do bad things, so we shouldn't even bother trying to do anything to make things better, and just let the dystopia happen" is not the answer