Anthropic’s showdown with the US Department of War may literally mean life or death—for all of us – Bulletin of the Atomic Scientists
SAFETY
-
Unreliability is Key, Not Imminent Superintelligence, Says Substack Author
By
–
or maybe you haven’t read actually my articles. i was perfectly plain in my substack yesterday (and many other essays and tweets) that the unreliability is the key, not superintelligence, which i have repeatedly said is not imminent. this one is on you, mate.
-
Generative AI Unreliability and Its Impact on Military Decisions
By
–
For all you fools who are lost, or can’t follow the complexity and nuance in my ideas, the following three facts are completely consistent with one another. 1. GenAI cannot be trusted because it is unreliable. 2. Using unreliable tech in high-stakes military decisions without x.com/chammond510/st…
-
Generative AI Unreliability and High-Stakes Military Decisions
By
–
For all you fools who are lost, or can’t follow the complexity and nuance in my ideas, the following three facts are completely consistent with one another. 1. GenAI cannot be trusted because it is unreliable. 2. Using unreliable tech in high-stakes military decisions without
-
Gary Marcus Shares Articles on AI’s Impact on Humanity and Society
By
–
wrote something similar in this https://
open.substack.com/pub/garymarcus
/p/code-red-for-humanity?r=8tdk6&utm_medium=ios
… and this https://
open.substack.com/pub/garymarcus
/p/america-and-probably-the-world-stands?r=8tdk6&utm_medium=ios
… -
AI Chatbot Hallucinations and Nuclear War: A Catastrophic Scenario
By
–
Sure, Meta lost 70 or 80 billion dollars chasing Palmer Luckey’s AR dreams, but if the Pentagon accidentally launches a nuclear war because some chatbot hallucinates, ain’t nobody gonna remember that.
-
AI Agents Need Sequential Modularity for Better Evaluation
By
–
2/4 Modern AI systems plan, retrieve across sources, reason, take actions, and update state. But most evals capture only the end result of that entire process. Agents need sequential modularity, like human language production, to enable evaluation per module and support
-
AI21 Labs focuses on frontier model capabilities failure taxonomy and agent validation
By
–
Thanks for this research & for building practical measures of frontier model capabilities. We're focusing these days on the failure taxonomy you highlighted in the paper – testing how elements like structured programmatic plans and in-flow validators & reducers can help agents
-
AI and Humanity: The Danger Lies in Human Misuse, Not AI Itself
By
–
You should not be afraid of AI exterminating the humanity. You should be afraid of the man who uses AI to exterminate the humanity.
-
Autonomous AI and Nuclear War: A Legal Gap?
By
–
*Generals* generally aren’t. And I will leave it at that. but note also that this bill died: https://
congress.gov/bill/118th-con
gress/senate-bill/1394
… and it is thus probably legal for autonomous AI to launch nuclear war.