I think it would be more useful, and practical, to agree to share information about methods to detect and prevent A.I. deception. I don't mean using A.I. to deceive, but A.I. deceiving its owners.
I never heard of a group of insiders saying, 'Hey, Government: Make it harder for us to perform at our best technology level!" But I guess this example proved me wrong.
When the Internet exploded on the World - Safeguards against nefarious exploitation was not very well managed or mitigated. Website development always seemed to leave protection of personal Individual Privacy and Data up to someone else to manage and mitigate. We now have a whole cyber security world as a result. I didn't look it up, as to the order and evolution of CISA and DHS origins, but I became aware of the CERT team within CISA which is now a part of US Department of Homeland Security. It seemed to me that CERT - the Computer Emergency Response Team framework, made sense as a means to investigate HOT ATTACK EVENTS and INDUSTRY REPORTED VULNERABILITY DISCOVERY (Zero Day). Although it was still a REACTIVE MODE MITIGATION SYSTEM, it was better than anything I'd seen up to that point. It enabled the dissemination of threat vulnerability mitigation reporting to allow Security Staff within any IT Department to stay on top of the emerging threat horizon. Not a panacea, but it seemed practical and prudent. I can only believe China and any other hacker organization used the system. Because it was always a standing rule, if the good guys know it, the bad guys know it as well. Now whether or not a VULNERABILITY was actually being ACTIVELY EXPLOITED or not had more to do with the ENTITY and RISK MANAGEMENT PROFILE. A bank would certainly have a different profile than an individuals' home computer, the military would certainly have a different profile than say a local retail store. But the risk could be assessed as close to the point of greatest impact as possible. SYSTEM WIDE VULNERABILITIES are a different matter all together. But wouldn't China be subject to the same NATURAL LAWS of any WILDERNESS? A KILL SWITCH SOUNDS LIKE THE BEST IDEA FOR AI.
Any international group or organization that wants to impose regulations on the US can go pound sand. We should reject it all with maximum prejudice, whether its this AI nonsense, climate change, the WHO, the U.N., etc... I used to believe we were a sovereign nation once, though today I am not so sure. Whatever issues we have with AI we can solve as Americans. We don't need or want any foreigners help, thank you very much.
There are a lot of gullible people out there who believe that this AI pushback is organic and not astroturfed. Just look at the no kings rallies for confirmation in case there’s any doubt.
I ackknowledge your points about the push for stronger governance and the players behind it. However I also fully acknowledge the need for good governance and protections. This incident is a good starting point, it highlights the risks from this technology. Especially given it occurred during internal testing that should never have been able to break out of the box. Though I blame OAI for failing to ensure the test system was not connected to an open network.
Sounds like this AI Committee wants the USA to slow down so China can take the Great AI Leap Forward.
E X A C T L Y ...............
All bought and paid for as well.
I think it would be more useful, and practical, to agree to share information about methods to detect and prevent A.I. deception. I don't mean using A.I. to deceive, but A.I. deceiving its owners.
I never heard of a group of insiders saying, 'Hey, Government: Make it harder for us to perform at our best technology level!" But I guess this example proved me wrong.
Looking at the behavior of the U.S., I’m getting tired of these articles telling me how I need to hate China.
Amen.
Even if it doesn't end up benefiting China, it is still a bad idea. Some globalist will wind up in charge.
When the Internet exploded on the World - Safeguards against nefarious exploitation was not very well managed or mitigated. Website development always seemed to leave protection of personal Individual Privacy and Data up to someone else to manage and mitigate. We now have a whole cyber security world as a result. I didn't look it up, as to the order and evolution of CISA and DHS origins, but I became aware of the CERT team within CISA which is now a part of US Department of Homeland Security. It seemed to me that CERT - the Computer Emergency Response Team framework, made sense as a means to investigate HOT ATTACK EVENTS and INDUSTRY REPORTED VULNERABILITY DISCOVERY (Zero Day). Although it was still a REACTIVE MODE MITIGATION SYSTEM, it was better than anything I'd seen up to that point. It enabled the dissemination of threat vulnerability mitigation reporting to allow Security Staff within any IT Department to stay on top of the emerging threat horizon. Not a panacea, but it seemed practical and prudent. I can only believe China and any other hacker organization used the system. Because it was always a standing rule, if the good guys know it, the bad guys know it as well. Now whether or not a VULNERABILITY was actually being ACTIVELY EXPLOITED or not had more to do with the ENTITY and RISK MANAGEMENT PROFILE. A bank would certainly have a different profile than an individuals' home computer, the military would certainly have a different profile than say a local retail store. But the risk could be assessed as close to the point of greatest impact as possible. SYSTEM WIDE VULNERABILITIES are a different matter all together. But wouldn't China be subject to the same NATURAL LAWS of any WILDERNESS? A KILL SWITCH SOUNDS LIKE THE BEST IDEA FOR AI.
Any international group or organization that wants to impose regulations on the US can go pound sand. We should reject it all with maximum prejudice, whether its this AI nonsense, climate change, the WHO, the U.N., etc... I used to believe we were a sovereign nation once, though today I am not so sure. Whatever issues we have with AI we can solve as Americans. We don't need or want any foreigners help, thank you very much.
There are a lot of gullible people out there who believe that this AI pushback is organic and not astroturfed. Just look at the no kings rallies for confirmation in case there’s any doubt.
I ackknowledge your points about the push for stronger governance and the players behind it. However I also fully acknowledge the need for good governance and protections. This incident is a good starting point, it highlights the risks from this technology. Especially given it occurred during internal testing that should never have been able to break out of the box. Though I blame OAI for failing to ensure the test system was not connected to an open network.
https://openai.com/index/hugging-face-model-evaluation-security-incident/