And the #AI can't distinguish whether it's #AI or real. https://
nytimes.com/interactive/20
23/06/28/technology/ai-detection-midjourney-stable-diffusion-dalle.html
… by @stuartathompson and @tiffkhsu
SAFETY
-
AI Cannot Distinguish Itself From Real Content
By
–
-
Securing AI Agents: Data Permissioning and Leak Prevention
By
–
We want to build AI agents that use internal data to create external interfaces (i.e. customer service bots). How should we be thinking about permissioning? How do we prevent agents from leaking internal data sources via prompt engineering or simply a mistake? @strandbrown
-
GPT-3 Generates More Convincing Disinformation Than Humans
By
–
A large language #AI model (GPT-3) produces more compelling disinformation than humans by generating information that is easier to understand https://
science.org/doi/10.1126/sc
iadv.adh1850
…
Humans can't distinguish between it and real human tweet posts @fedgermani @UZH_en @culturico -
Cyber Resilience for Energy Operations Safety and Continuity
By
–
Learn the essential considerations for developing cyber resilience and ensuring your energy operations' safety, reliability, and continuity. Join the @Fortinet LIVE session at #IIoTWorldDay! https://
ow.ly/uzjA50OSKOt #sponsored #fortinet_ics #cybersecurity #energyresilience #energy -
AI Reliability Issues and Performance Benchmarking Improvements
By
–
Plenty of room for improvement here. Sometimes it behaves like an actual AI with ability to handle 20 follow ups, sometimes just like someone with schizophrenia. We are working on benchmarks and improvements.
-
Extinction Risk vs Catastrophic Risk in AI Safety
By
–
Extinction risk is not the same as catastrophic risk; this is a good paper, primarily on the latter.
-
Distinguishing Catastrophic Risk from Extinction Risk in AI
By
–
This is a good paper, and rightly distinguishes catastrophic risk from extinction risk.
-
Google Cloud CEO Discusses AI Risks with EU Regulators
By
–
New: in an exclusive interview, the CEO of @googlecloud acknowledged the risks of AI and told me Google is having "productive" early talks with EU officials. "We do want to find a path forward,” Kurian told me at the company's Victoria offices yesterday.
-
Human Editorial Content as Defense Against AI-Generated Content
By
–
I asked GPT to respond to you: GPT: “Rafat's tweet implies that high-quality editorial content crafted by humans stands as a robust defense in this era of rapidly evolving AI-generated content. This perspective mainly rests on the notion that human creativity, in-depth emotional
-
Pragmatism Over Speculation: Addressing Today’s AI Risks
By
–
Finally, some pragmatism. “Stop talking about tomorrow’s AI doomsday when AI poses risks today.”