Adversarial Policies Beat Superhuman Go AIs Their policies do not win by playing Go well. Instead, they trick AI into making serious blunders! These tricks transfer to other superhuman Go AIs and can be used by humans to consistently beat superhuman AIs. https://
arxiv.org/abs/2211.00241
SAFETY
-
Adversarial Policies Exploit Weaknesses in Superhuman Go AIs
By
–
-
AI Watermarking Promises Fall Short on Synthetic Text Pollution
By
–
The inclusion of watermarking is welcome, of course, but I bet that falls under "we promise to start working on this" and not "we'll stop polluting the information ecosystem with extruded synthetic text until we have it".
-
Celebrities Embrace Artificial Intelligence Deepfakes Technology
By
–
Why some celebrities are embracing Artificial Intelligence deepfakes. #ChatGPT #OpenAI #GenerativeAI #AI #ArtificialIntelligence #MachineLearning #DeepLearning #Tech #Innovation #IoT #TechNews https://
bbc.com/news/business-
65995089
… -
World Ethical Data Foundation releases 84-question AI safety guidelines
By
–
AI: Experts propose guidelines for safe systems. The World Ethical Data Foundation has 25,000 members including staff from #Meta, #Google and #Samsung. Framework contains 84 Qs for #developers at the start of an #AI project. #ArtificialIntelligence #IoT
-
Australian Team Wins $600k Grant for AI-Brain Cell Integration
By
–
Australian DishBrain team wins $600,000 grant to merge AI with human brain cells despite concerns. #ChatGPT #Google #OpenAI #GenerativeAI #AI #ArtificialIntelligence #MachineLearning #Tech #Innovation #ML #IoT #TechNews #DeepLearning
-
AI Companies Establish Guardrails on Emerging Tools
By
–
#AI Companies Agree to Guardrails on New Tools https://
nytimes.com/2023/07/21/us/
politics/ai-regulation-biden.html?smid=tw-share
… -
OODAcon 2023 Conference on Global Risk and Cybersecurity
By
–
OODAcon 2023 is about the future, now. Track topics and get more info at: http://
oodaloop.com/oodacon-2023/ #globalrisk #cybersecurity #nationalsecurity -
Notkilleveryoneism: AI as Unworthy Heir to Humanity
By
–
Similarly, be aware that notkilleveryoneism is not "biological humans should be on top of Existence forever, because carbon chauvinism". Notkilleveryoneism is "We predict that, at the current rate, AI will kill everyone because it will not be a worthy heir to humanity."
-
Health and Longevity Matter, But What Comes After
By
–
To be explicit, I'd very much accept human health and longevity as a preferable alternative to our total extermination and replacement by squiggle-maximizers. But what happens after the health and longevity does matter.
-
AI surpassing human cognition requires guardrails and technical advances
By
–
“At some point over the next decade, without real guardrails and new technical advances, it's possible that an AI might quickly get better than us at every conceivable cognitive task.”