AI Industry Struggles to Curb Misuse as Users Exploit Generative AI for Chaos::Artificial intelligence just can’t keep up with the human desire to see boobs and 9/11 memes, no matter how strong the guardrails are.
This is a most excellent place for technology news and articles.
From the spirits that I called, Sir, deliver me!
😁
Why didnt someone warn us about this? Nobody said this might happen, nobody! Not a single person tried to be the voice of reason!
Is this really something people are mad about? Who cares? This shit is hilarious.
No, no one cares but it gets a bunch of clicks because it’s hilarious so articles keep getting written.
It’s a solved problem too. You just run the prompt and the result of the generation through a second pass of a fine tuned model checking for jailbreaking or rule breaking content generation.
But that increases cost per query by 2-3x.
And as you said, no one really cares, so it’s not deemed worth it.
Yet the clicks keep coming in for anti-AI articles, so they keep getting pumped out, and laypeople now somehow think jailbreaking or hallucinations are intractable problems preventing enterprise adoption of LLMs, which is only true for the most basic plug and play high volume integrations.
Huh, so basically it’s like every time my mom said “think before you speak”. You know, just run that line in your head once before you actually say it, to avoid saying something dumb/offensive.
Of all the fucking things to worry about with AI… Pregnant sonic being behind 9/11.
Well I mean it points to our inability to control the use of ai systems, that is in fact a very real problem.
If you can’t keep people from making stupid memes, you also can’t keep people from making misleading propaganda or other seriously problematic content.
Towards the end of the story there was the example where they couldn’t stop the system from giving people a recipe for napalm, despite “weapons development” being an explicitly banned topic. I don’t think I need to spell out how that’s a problem.
Was not expecting to see a pregnant sonic flying a plane today.
And Ganondorf is the father
This is a part of a bigger topic people need to be aware of. As more and more AI is used in public spaces and the internet, people will find creative ways to exploit it.
There will always be ways to make the AI do stuff the owners don’t want it to. You could think of it like the exploits used in speedrunning, but in this case there’s a lot more variety. Just like you can make an AI generate morally questionable material, you could potentially find a way to exploit the AI of a self driving car to do whatever you can think of.
This is trivially fixable, it’s just at 2-3x the per query cost so it isn’t deemed worth it for high volume chatbots given the low impact of jailbreaking.
For anything where jailbreaking would somehow be a safety concern, that cost just needs to be factored in.
That’s true for all the things that can have a query cost. What about those AI applications that don’t have any financial cost to the user? For instance, The Spiffing Brit continues to find interesting ways to exploit the YouTube Algoritm. I’m sure you can apply that same “hacker mentality” to anything with AI in it.
At the moment, many of those applications are on the web, and that’s exactly where a query costs can be a feasible way to limit the number of experiments you can reasonably run in order to find your favorite exploit. If it’s too expensive, you probably won’t find anything worth exploiting, and that should keep the system relatively safe. However, nowadays more and more AI is finding its way in the real world, which means that those exploits are going to have some very spicy rewards.
Just imagine if the traffic lights were controlled by an AI, and you found an exploit that allowed you to get the green light on demand? Applications like this don’t have any API query costs. You just need to be patient and try all sorts of weird stuff to see how the lights react. Sure, you can’t run a gazillion experiments in an hour, which means that you might not find anything worth exploiting. Since there would be millions of people experimenting with the system simultaneously, surely someone would find an exploit.
Serious question - why should anyone care about using AI to make 9/11 memes? Boobs I can see the potential argument against at least (deep fakes and whatnot), but bad taste jokes?
Are these image generation companies actually concerned they’ll be sued because someone used their platform to make an image in bad taste? Even if such a thing we’re possible, wouldn’t the responsibility be on the person who made it? Or at worst the platform that distributed the images -As opposed to the one that privately made it?
Protect the brand. That’s it.
Microsoft doesn’t want non-PC stuff being associated with the Bing brand.
It’s what a ton of the ‘safety’ alignment work is about.
This generation of models doesn’t pose any actual threat of hostile actions. The “GPT-4 lied and said it was human to try to buy chemical weapons” in the safety paper at release was comical if you read the full transcript.
But they pose a great deal of risk to brand control.
Yet still apparently not enough to run results through additional passes which fixes 99% of all these issues, just at 2-3x the cost.
It’s part of why articles like these are ridiculous. It’s broadly a solved problem, it’s just the cost/benefit of the solution isn’t enough to justify it because (a) these issues are low impact and don’t really matter for 98% of the audience, and (b) the robust fix is way more costly than the low hanging fruit chatbot applications can justify.
I’d guess that they are worried the IP owners will sue them for singing their IP.
So sonic creators will say, your profiting by using sonic and not paying us for the right to use him.
But I agree that deep fakes can be pretty bad.
You are profiting = you’re profiting.
I don’t see adobe trying to stop people from making 911 memes in photoshop nor have they been sued over anything like that, I dont get why AI should be different. It’s just a tool.
The problem for Adobe is that the AI work is being done on their computers, not yours, so it could be argued that they are liable for generated content. ‘Could’ because it’s far from established but you can imagine how nervous this all must make their lawyers.
That’s a great analogy, wish I’d thought of it
I guess it comes down to whether the courts decide to view AI as a tool like photoshop, or a service - like an art commission. I think it should be the former, but I wouldn’t be at all surprised if the dinosaurs in the US gov think it’s the latter
Welcome to the Internet
who could’ve seen this coming
I busted out laughing on a public bus while reading grandma’s napalm recipe.
“Karl Marx large breasts” got me
In Australia we are currently voting on a constitutional amendment. It would create an advisory body that represents first nations people. It’s super basic, doesn’t really cover how it will work, because they can’t really even work on that until the amendment passes.
But presumably it will allow them to directly advise government, rather than through the spiderweb of community leaders, NGOs and whatnot that exist now, and provide some structure for Aboriginal representation in parliament
The sheer amount of disinformation circulating is staggering. I was lucky enough to really avoid most of the drama, until I went and had a look this past week finally.
What interested me, was rather than the usual short posts and snarky racist comments, of which plenty exist. These long diatribes were dominant, on places like Reddit and Facebook.
Then it struck me, they all sound like they were written by the same person. Not just a little, if you had removed the names and pictures of the users, I would have flat out assumed it was the same person.
We have opened Pandora’s Box. We don’t need “AGI” or whatever, this is plenty enough to do us in.