| AI and terrorism 10:59 - Sep 11 with 813 views | Swansea_Blue |
Apparently Anthropic have prevented Claude being used to develop bioweapons - https://www.bbc.co.uk/news/art
This is from the same company that was brought up at PMQ’s this week:
(buried in BBC live reporting of PMQs)
“Pushed on AI safety, Burnham invites cross-party discussion
published at 12:31 9 September
Davey says the best legacy of the last Conservative PM Rishi Sunak is the UK's "world-leading" AI institute. But today, the FT newspaper is reporting tech firm Anthropic did not submit its latest model for testing, apparently due to pressure from the Trump administration.
He asks if the PM thinks it's acceptable for the US president to undermine efforts to keep the world safe from dangerous AI.
Burnham says he has appointed a minister for AI in cabinet and boosted the conversation of AI in government. He says it is open to MPs on all sides to be part of the discussion.”
Shouldn’t we be insisting companies meet any of our requirements for testing and banning their products if we don’t comply. The security implications here are huge. Should we not be regulating the crap out of products that can be used to attack the UK? |  |
| |  |
| AI and terrorism on 11:03 - Sep 11 with 775 views | Dubtractor |
I know I'm a bit of a luddite, but the extent to which AI is becoming embedded in absolutely everything, with seemingly little oversight or regulation, feels like a massive recipe for disaster.
I've used the comparison before, but the terminator films were supposed to be speculative fiction, not templates to follow. |  |
| |
| AI and terrorism on 11:11 - Sep 11 with 731 views | Swansea_Blue |
| AI and terrorism on 11:03 - Sep 11 by Dubtractor |
I know I'm a bit of a luddite, but the extent to which AI is becoming embedded in absolutely everything, with seemingly little oversight or regulation, feels like a massive recipe for disaster.
I've used the comparison before, but the terminator films were supposed to be speculative fiction, not templates to follow. |
Something popped up yesterday about it increasingly being used to write stuff in government business (I can’t remember what exactly). It was significant 25-30%ish now being written using AI; briefs or reports of whatever it was. It seems everywhere.
I’ve found it useful at a personal level for planning and researching. A bit of a game changer really, so I can see it becoming the main thing we talk to. It seems wired to pander to us too, so addictive and I can see it replacing human contact for some people. But it gets so much wrong too (when it comes to internet searches ). Confidently wrong. ChatGPT was adamant in telling me a 3 year old laptop that once had good reviews was still a best buy when I was looking yesterday. Minor stuff compared to us all dying in a BioWare of course! [Post edited 11 Sep 11:15]
| |
| |
| AI and terrorism on 11:27 - Sep 11 with 677 views | Guthrum |
AI is no more or less useful to terrorism than Google. It cannot, on its own, develop bioweapons (or explosives, certainly not nuclear) in an insurgent context. There would still need to be labs, technicians and supply chains. There would still need to be agent networks for covert distribution and use. At best, it would be a research tool, at worst, wildly misleading if the information sought is not available and the AI makes it up.
The one area where AI might be of use to terrorists is in the realm of hacking, where speed of operation (e.g. cracking passwords) could give an advantage. However, it's also the area where security agencies and commercial departments have been devoting an awful lot of work to counter that.
AI is most dangerous when placed, unsupervised, in charge of important systems, which then run off the rails due to rigid but unsophisticated logic or simple errors it doesn't know are wrong. |  |
| |
| AI and terrorism on 11:40 - Sep 11 with 622 views | DanTheMan |
| AI and terrorism on 11:27 - Sep 11 by Guthrum |
AI is no more or less useful to terrorism than Google. It cannot, on its own, develop bioweapons (or explosives, certainly not nuclear) in an insurgent context. There would still need to be labs, technicians and supply chains. There would still need to be agent networks for covert distribution and use. At best, it would be a research tool, at worst, wildly misleading if the information sought is not available and the AI makes it up.
The one area where AI might be of use to terrorists is in the realm of hacking, where speed of operation (e.g. cracking passwords) could give an advantage. However, it's also the area where security agencies and commercial departments have been devoting an awful lot of work to counter that.
AI is most dangerous when placed, unsupervised, in charge of important systems, which then run off the rails due to rigid but unsophisticated logic or simple errors it doesn't know are wrong. |
"where speed of operation (e.g. cracking passwords) could give an advantage."
LLMs would be useless at this, for what it's worth. Password cracking is basically down to how much compute power you can throw at things, it's mostly maths. |  |
| |
| AI and terrorism on 11:43 - Sep 11 with 607 views | Kievthegreat |
| AI and terrorism on 11:40 - Sep 11 by DanTheMan |
"where speed of operation (e.g. cracking passwords) could give an advantage."
LLMs would be useless at this, for what it's worth. Password cracking is basically down to how much compute power you can throw at things, it's mostly maths. |
Bigger risk for password cracking and encryption/decryption is surely usable quantum computing? |  | | |
| AI and terrorism on 11:45 - Sep 11 with 597 views | Guthrum |
| AI and terrorism on 11:40 - Sep 11 by DanTheMan |
"where speed of operation (e.g. cracking passwords) could give an advantage."
LLMs would be useless at this, for what it's worth. Password cracking is basically down to how much compute power you can throw at things, it's mostly maths. |
OK, so AI isn't even much use for that, then. | |
| |
| AI and terrorism on 11:49 - Sep 11 with 588 views | Guthrum |
| AI and terrorism on 11:27 - Sep 11 by Guthrum |
AI is no more or less useful to terrorism than Google. It cannot, on its own, develop bioweapons (or explosives, certainly not nuclear) in an insurgent context. There would still need to be labs, technicians and supply chains. There would still need to be agent networks for covert distribution and use. At best, it would be a research tool, at worst, wildly misleading if the information sought is not available and the AI makes it up.
The one area where AI might be of use to terrorists is in the realm of hacking, where speed of operation (e.g. cracking passwords) could give an advantage. However, it's also the area where security agencies and commercial departments have been devoting an awful lot of work to counter that.
AI is most dangerous when placed, unsupervised, in charge of important systems, which then run off the rails due to rigid but unsophisticated logic or simple errors it doesn't know are wrong. |
There is another danger to civilisation posed by AI - that the mush it is allowed to generate when not tightly instructed and trained will rot humanity's collective intelligence. It will further blur the distinction between reality and made up/misleading stuff. | |
| |
| AI and terrorism on 11:54 - Sep 11 with 569 views | DanTheMan |
| AI and terrorism on 11:43 - Sep 11 by Kievthegreat |
Bigger risk for password cracking and encryption/decryption is surely usable quantum computing? |
Password cracking, probably not, at least with modern hashing.
Encryption, possibly for state actors, but not something the average user should be worrying about at least.
https://www.gopher.security/po
Article is a bit alarmist but that's the gist of it.
We do already have post-quantum cryptography in browsers for example.
https://en.wikipedia.org/wiki/ | |
| |
Login to get fewer ads
| AI and terrorism on 12:21 - Sep 11 with 504 views | homer_123 |
| AI and terrorism on 11:49 - Sep 11 by Guthrum |
There is another danger to civilisation posed by AI - that the mush it is allowed to generate when not tightly instructed and trained will rot humanity's collective intelligence. It will further blur the distinction between reality and made up/misleading stuff. |
Basically, it can ruin us with mis-information and mis-direction. Though we don't really need any help with that. It could stand back and watch... |  |
| |
| AI and terrorism on 13:03 - Sep 11 with 422 views | Swansea_Blue |
| AI and terrorism on 11:49 - Sep 11 by Guthrum |
There is another danger to civilisation posed by AI - that the mush it is allowed to generate when not tightly instructed and trained will rot humanity's collective intelligence. It will further blur the distinction between reality and made up/misleading stuff. |
I think that’s the big one. In a digital online landscape already marred by disinformation the scope for bad actors to pump out realistic looking propaganda has jumped up a notch. That ultimately threatens democracy, not just the dumbing down of human intelligence. | |
| |
| AI and terrorism on 13:47 - Sep 11 with 317 views | Illinoisblue |
This makes me recall the days when BBC and ITV had to use actor’s voices to carry the words of IRA leaders. Seems almost quaint. |  |
| |
| AI and terrorism on 16:45 - Sep 11 with 234 views | WicklowBlue |
| AI and terrorism on 13:03 - Sep 11 by Swansea_Blue |
I think that’s the big one. In a digital online landscape already marred by disinformation the scope for bad actors to pump out realistic looking propaganda has jumped up a notch. That ultimately threatens democracy, not just the dumbing down of human intelligence. |
Indeed look what the Turnip in the Whitehouse has been churning out with AI! |  | | |
| |
By continuing to use the site, you agree to our use of cookies and to abide by our Terms and Conditions.
