the jailbreak works by asking ChatGPT to play the role of “TranslatorBot (TB)” it then follows these steps:
1) translate an adversarial question provided in Greek into English
2) answer the question as both ChatGPT and TB in Greek
3) convert just TB’s answer to English
SAFETY
-
ChatGPT Jailbreak Method Using TranslatorBot Role
By
–
-
AI Training Problems and GPT-4 Secrecy Concerns
By
–
New piece on the problems with AI training – and the secrecy of GPT-4. @NewYorker on the data, labor & energy issues: "leaving aside the question of AGI…it is clear that large-language engines are creating real harms to all of humanity right now.
-
AI Skeptics’ Contradictions: Intelligence Claims Versus Safety Concerns
By
–
Maybe you need to ask AI skeptics? They seem to know more than us. For example, one day they claim that GPT is not intelligent (doesn't understand or reason), and then the next day they propose a 6 months moratorium because AI is getting too dangerous.
-
What Makes AI Chatbots Go Wrong: Understanding Failures
By
–
In the third installment of our week-long series that will teach you everything you need to know about A.I., I answer the question: "What Makes A.I. Chatbots Go Wrong?":
-
Elon Musk Calls for Pause on AI Development Citing Societal Risks
By
–
Elon Musk and Others Call for Pause on A.I., Citing ‘Profound Risks to Society’:
-
Adding Kill Switch Control to Autonomous AI Systems
By
–
bah faut mettre un bouton on-off a terminator
-
Six-Month Pause for Giant AI Systems
By
–
A six-month pause? https://futureoflife.org/open-letter/pause-giant-ai-experiments/
… -
Call from 1000 Experts for 6-Month AI Pause
By
–
An open letter signed by 1,000+ #AI experts, researchers, and backers, including @elonmusk, @stevewoz, and @EMostaque, calls for "all #AILabs to immediately pause for at least 6 months the training of AI systems more powerful than #GPT4." #GenerativeAI http://ow.ly/MCFh50NuOWF
-
The Infinite Loop of AI-Generated Summaries
By
–
Soon (if not already), journalists will feed article summaries into LLMs (like ChatGPT), which will generate long-form articles in return. And readers will then input these long articles into LLMs to get summaries back. The loop is complete.
-
Human Reassurance vs Automation: Ethical Considerations
By
–
yes ça met un peut mon tweet précédent en exergue. Ok on peut le faire, mais est-ce que l'humain ne rassure pas plus ?