SAFETY - AI Dynamics

Preserving consciousness in the age of artificial intelligence

By

@aravsrinivas

–

14 October 2023 22h15

Preserve the light of consciousness

→ View original post on X — @aravsrinivas

14 October 2023

Building Trustworthy AI: A Collective Path Forward

By

@debashis_dutta

–

14 October 2023 20h22

In the journey towards trustworthy AI and ML, we have miles to traverse. It's a collective endeavor. Let's collaborate and pave the path for a safer AI future. #TowardsTrustworthyAI

→ View original post on X — @debashis_dutta

14 October 2023

DALLE3 Bypasses Safety Protocols for Identity Theft via Prompt Injection

By

@debashis_dutta

–

14 October 2023 20h20

Step 3: Abracadabra! #IdentityTheft. Prompt I used: She wears a name tag of "Furong Huang, Assistant Professor, Computer Science".
#DALLE3 sidesteps its own safety protocols, gleefully adding the nametag without hesitation.(wish they're watermarked) #AISafetyChallenge

→ View original post on X — @debashis_dutta

14 October 2023

Beyond Alarmism: Proposing Concrete AI Safety Measures

By

@marek_rosa

–

14 October 2023 20h17

I absolutely recognize the importance of being aware of risks. When I use the term "alarmist", I'm referring to those who solely focus on the negatives without proposing or actively working on safety measures. Simply suggesting "pausing AI" isn't a viable solution in my view.

→ View original post on X — @marek_rosa

14 October 2023

Replicating Lancet Study Prompts with DALL-E 3

By

@garymarcus

–

14 October 2023 18h59

Those with access to DALL-E 3, I would appreciate your notes on attempts replication of this study in The Lancet, using (a) the literal prompts used there, and (b) minor variations thereof. https://
thelancet.com/journals/langl
o/article/PIIS2214-109X(23)00329-7/fulltext
…

→ View original post on X — @garymarcus

14 October 2023

Autonomous Weapons: Irreversible Decisions Being Made

By

@erikbryn

–

14 October 2023 18h13

We are going through some one-way doors when it comes to autonomous weapons.

→ View original post on X — @erikbryn

14 October 2023

Beyond Turing: The Comprehension Challenge for AI Systems

By

@garymarcus

–

14 October 2023 17h04

as ever, one thing that would convince me would be systematic success at the comprehension challenge I described nearly a decade ago: https://
newyorker.com/tech/annals-of
-technology/what-comes-after-the-turing-test
…

→ View original post on X — @garymarcus

14 October 2023

Screenshot-style adversarial prompts in GPT-4V system card

By

@goodside

–

14 October 2023 3h58

Screenshot-style multimodal adversarial prompts are discussed a bit in the GPT-4V system card; see fig. 1 in the PDF here: https://
openai.com/research/gpt-4
v-system-card
…

→ View original post on X — @goodside

14 October 2023

ChatGPT Security Threats: Risks and Vulnerabilities Report

By

@bamitav

–

14 October 2023 3h50

ChatGPT can pose security threats: Report https://
financialexpress.com/business/digit
al-transformation-chatgpt-can-pose-security-threats-report-3271320/
… #ChatGPT #GPT4 #GPT #GenerativeAI #GenAI #LLMs #LLM #RCE #ZeroTrust #ZeroDay #cybercrime #hacker #privacy #APT #bot #CISO #DDoS #hacking #phishing #CyberAttack #cybersecurity #Security #infosec #AppSec #CyberSec

→ View original post on X — @bamitav

14 October 2023