I get that AI is trending as the big scary technology right now. Sure, there are issues.
But can everybody in the news just stop, take a deep breath, and fucking learn something about AI before panic-reporting the latest non-event from an AI platform. At least learn enough about it to ask intelligent questions and not fall for every 20-something self-professed AI “expert” from the marketing department telling us humanity is doomed in a matter of weeks if all AI isn’t immediately killed with fire.
And for fuck-sake, stop saying ‘sandbox’ like you know what a sandbox is.
I mean, BBC, NPR, CNN, Youtube feeds. It’s unavoidable.
I’m in IT and I build AI agent solutions (not willingly), and even our security guy is sending me Claud generated warnings of OpenAI security vulnerabilities. It’s all become recursively perverse.
It’s fed also by Doom trolling: companies cough OpenAI / Sam Altman cough that want to inflate the value of their tools by “warning” of how powerful they are.
A lot of news is lazily repeating press releases. That part isn’t new. But now companies are incentivized to have chicken little write those press releases.
News is about eyeballs/ears, ratings, advertisement, and money. They are always looking for the next big hot story to talk about, and the elimination of humanity in 10 years sounds like a pretty hot story to me. Whether that story is accurate or not has nothing to do with it. All that matters is whether it will “sell” to a mass naive, audience.
I disagree. Anyone here feel a little panicked about climate change? Well AI is changing 100 times as fast as the climate.
While I certainly think many of the problems of AI could be fixed by intelligent, rational people with decent moral values; these type people are not in control of either the AI industry nor the government.
Surely you’re not referring to the top AI researchers and executives (the ones issuing a lot of the dire warnings I’ve seen lately) as “20-something self-professed AI ‘experts’ from the marketing department?”
I avoid all of those things. It’s not unavoidable.
You choose to immerse yourself in performative alarm-mongering and then wonder why you’re alarmed. You’re welcome to do whatever you want and I wish you well.
But we each can choose our actions or our consequences. Not both.
Think of a sandbox as a controlled playground where applications, code, and files can be tested or executed to see how they behave. If the software behaves maliciously or unexpectedly, it doesn’t have the power to affect anything outside of that contained environment.
So say AI experts.
It’s a good thing they’re not zoo keepers or we’d be up to our asses in monkeys.
And specifically one AI thing supposedly broke through the sandbox, which they present as “look at how powerful our AI has gotten!” But the scuttlebutt is that the sandbox was not designed well, and they wanted this to happen.
The actual fear is not unfounded: an AI that can basically create malware to keep itself running even after you try to shut it down. The malware would basically copy the AI somewhere else, where it keeps copying itself and mutating. Basically a worm that can mutate faster than humans can come up with countermeasures to stop it.
But worms (even mutating ones) spread in large part because the payload is quite small, and AI models only get sophisticated when they get rather big.
