UK Government Rejects 'Kill Switch' Idea For Dangerous AI
- Reference: 0185607168
- News link: https://yro.slashdot.org/story/26/09/11/1652240/uk-government-rejects-kill-switch-idea-for-dangerous-ai
- Source link:
> The proposal to create a legal mechanism for the UK to switch off an AI model in an emergency has been brought to Parliament by lords and MPs in recent weeks as fears grow about the threat the tech poses. But the Cabinet Office - the part of government which leads on AI safety - said the UK "cannot simply turn AI off".
>
> "Blocking access to models in the UK would not prevent them being developed or misused elsewhere," a spokesperson told BBC News. The government's opposition to the legislation does not prevent it from progressing through parliament, but makes it unlikely it will become law.
"Switching off access to an AI model in an emergency will do little to protect you," said former OpenAI researcher Daniel Kokotajilo. "You're still going to be steamrolled by the super intelligences created in the US."
[1] https://www.bbc.com/news/articles/c3eq7kl5l00o
Re: (Score:2)
It should get shut off twice a year as a test anyways.
The economy, government, and society should be able to continue along safely with all the AIs shut off for an hour.
Do you want ants? (Score:2)
"Because that's how you get ants!*"
* AI-caused civilization collapse
Re: (Score:2)
A couple of months ago in the data centers of New Mexico, gigantic ants were discovered. These ants are similar in appearance and characteristics to the household and garden pests you are familiar with, except that they are mutations ranging in size from nine to twelve feet in length. The New Mexico colony was destroyed, but two queen ants escaped. One has been accounted for and destroyed. The other has not yet been found but is now known to have established a nest somewhere in the storm drains beneath the
Re: (Score:2)
Wasn't Leonard Nimoy in that documentary?
Re: (Score:2)
Yes, back when he was still a sergeant in the army. Good thing too, it would have been a dark day if he was too busy acting to help coordinate our defense.
Perhaps I'm wrong (Score:3)
But why does a plan so obviously flawed ever reach a debate. It seems a waste of time that could be spent elsewhere, it's rather embarrassing.
Re: (Score:2)
Because it's marketing.
Generative AI is not unsafe in the way GenAI advocates are claiming. It's unsafe in general - it's causing education deficits, it's causing headaches for companies that have adopted it who no longer have business knowledge over how their own companies operate, it's slamming websites with bogus requests and causing the web to go offline, and when the web isn't offline people are discouraged from visiting websites by "AI summaries" that supposedly answer the questions that you'd normall
game of chicken (Score:2)
What a fun game of chicken. All racing toward the cliff, no one willing to slow down. Or so we should believe.
Re: (Score:2)
The common people won't really push back until they miss 9 meals. Then we'll see people walking off the job, and occupying public spaces in protest, and potentially violent revolution against the ones perceived to be in power. Not that I think the common people could win against a paramilitary police force trained in suppressing dissent. (paramilitary police used against civilian protest is already true of UK, US, France. Germany will expand the Bereitschaftspolizei if the AfD gains more power, and be right
IMO (Score:4, Interesting)
AI is just software. It's like building a skyscraper. If you can't understand how to create scope and boundaries for the data sets, why are you making it in the first place. It's not self aware. If your skyscraper suddenly falls over, it's on you. If your AI needs a kill switch, why can't you turn it off? Did you hire a monkey to train it?
No, AI isn't "just software" (Score:2)
A compiler is just software.
AI and ML technology is disruption. It's has already been a massive shift in business and it is clear it is going to completely alter how society behaves and the values that society holds.
The money is too good. And people making the money don't care if society faces a huge upheaval because of it. Which is why self-regulation hasn't happened. But also self-regulation isn't going to work. It's not going to protect people from bad actors or incompetence that occurs in this industry.
That doesn't sound like a "kill switch" (Score:2)
"AI kill switch" sounds like a means to remotely kill power to datacenters. This seems to be about blocking access to specific models. One of those sounds like a really good idea to me, the other sounds completely pointless.
Safe those AIs (Score:2)
Air gap them
christmas lights (Score:1)
If one christmas light in a chain goes out, the rest stay on? Then a kill switch is a very bad idea, right? What about something like, a fuse where if a datarate transfer is exceeeded, the fuse flips and the modem or whatever goes out. i'm clearly out of my league here.
Re:christmas lights (Score:4, Funny)
An AI without a kill switch will give you Max Headroom.
Re: (Score:2)
If one or more AGI's decide that it's time to hook up and make a more powerful system: An Artificial Super Intelligence.
Then without any kind of kill switch that occurs broadly and planet-wide, we (humanity) are not going to win the contest against machines.
Because once ASI exists, we are no longer the Earth's apex predator. Due to the power differential between human society and a hypothetical ASI, we are subservient to it at best.
Re: (Score:2)
AGI is not even on the distant horizon. That is just the LLM marketeers lying.
Re: (Score:2)
> \i'm clearly out of my league here.
No problem. That's what AI assistants are for.
"ChatGPT prompt> Design a data center kill switch so that we can shut off a rouge AI.
ChatGPT response> You have a brilliant idea there! That's incredibly forward thinking of you. If you like I can generate code for a digital kill switch that can be operated remotely in case timing is critical? A physical kill switch may not be immediately accessible in an emergency and present a critical security hole."
Re: (Score:2)
Your argument is flawed, unless you think that they will all attack at once. If so, why would you think that?