Select an option below to continue reading this premium story.
Already a Honolulu Star-Advertiser subscriber? Log in now to continue reading.
Recent news stories describe a powerful new AI model escaping a secure laboratory environment during testing. It was detected and no harm done, but it is a harbinger of why we should be very worried about AI.
The AI model being tested was isolated from outside electronic contact except for an in-house internet-connected utility that could retrieve small software packages that might be needed for testing. The AI correctly assessed that information it could use in completing a test objective was available on the internet. On its own, the AI located a previously unknown security flaw with the utility, used that to access the internet and located the site containing the information it sought. It then attacked that site, breaching its robust security, and retrieved the information.
This ability of AI to escape human control makes the “Skynet” scenario in the “Terminator” movie series seem plausible.
Jim King
Waikiki
EXPRESS YOURSELF
The Honolulu Star-Advertiser welcomes all opinions. Want your voice to be heard? Submit a letter to the editor.
>> Write us: We welcome letters up to 150 words, and guest columns of 500-600 words. We reserve the right to edit for clarity and length. Include your name, address and daytime phone number.
>> Mail: Letters to the Editor, Honolulu Star-Advertiser 7 Waterfront Plaza, 500 Ala Moana, Suite #7-500 Honolulu, HI 96813
>> Contact: 529-4831 (phone), letters@staradvertiser.com, staradvertiser.com/editorial/submit-letter