Many industry-leading AI companies have weaker safety guardrails than you think... today Haize Labs released some crazy jailbreaks they did on popular LLMs, ones you and your kids probably use. The results (images, videos, audio) are quite disturbing. See more in The Washington Post and on Twitter (linked below). Lucky to be a small investor, great work Leonard Tang! > https://lnkd.in/gNzdBPDu > https://lnkd.in/gUCgJSDA
Sara Du’s Post
More Relevant Posts
-
AI dangers, fake nude images, and sextortion concerns discussed | Flipboard: Artificial intelligence poses a growing threat, as it has been used to create and distribute fake explicit images of minors. FBI Agent Matthew Fowler shares insights on safeguarding our children against this issue. - Artificial Intelligence topics! #ai #artificialintelligence #intelligenzaartificiale
AI dangers, fake nude images, and sextortion concerns discussed | Flipboard
flipboard.com
To view or add a comment, sign in
-
Founder of SIFOR - Redefining Games with AI Innovation || AI & Games || Ludo AI enthusiast || 9 years in the mobile games scene
YouTube Lets You Beg for Deepfake Removal: Protecting Your Face, One Click at a Time Users can report videos that make them sound or look like they're doing something they never did... https://buff.ly/3RSauSY #tech #technology #news #latest #update #YouTube #ai
To view or add a comment, sign in
-
AI is not the boogeyman. It will help so e of those scammers out there. This means we need to be vigilant and security minded. Here is an article by another of the world’s best and brightest debunking the AI as Terminator myth.
A theoretical physicist says AI is just a ‘glorified tape recorder’ and people’s fears about it are overblown
businessinsider.com
To view or add a comment, sign in
-
Marketing Manager & AI Strategist | Expert in Technology Innovation & AI Solutions | Creative AI-Driven Design | Co-Founder and Manager at Doukani LLC
Are deepfakes the new frontier for misinformation or an untapped creative tool? The rise of deepfake technology presents a new wave of challenges, blurring the lines between reality and fabrication. Imagine a world where seeing is no longer believing. How do we trust what we see online when AI can convincingly alter video and audio? Advances in AI bring both risks and rewards. While deepfakes can innovate in film and education, they also pose serious threats to privacy and truth. Regulation and technology must evolve together to detect and manage deepfake content, ensuring it's used ethically and responsibly. What's your take on deepfakes—dangerous deception or a revolutionary tool? Share your thoughts. #Deepfake #AIethics #DigitalTrust #TechInnovation #ArtificialIntelligence
To view or add a comment, sign in
-
Mixing Code and Culture for a Brighter Tomorrow! 💻✨: Lead Technology & Diversity Specialist at Mont-Ford.
The power of AI strikes again, I am sure you have all seen this trend taking social media by storm ☁ Imagine being able to take that gorgeous sunset you snapped last summer, and make it wider, you know, make it even more postcard-perfect. And not just any old stretch - we're talking clever autofill that keeps everything in context. It's pretty simple really - you click, the AI does its magic, and voila! Your image is taken to a whole new level. After all, the beauty's in the details, right? But ethically could this have implications? We have to consider: Misleading Viusal Representation Privacy Violations Forgery & Fraud Deep Fakes Manipulation How do you feel about granting public access to such software without well-defined guidelines? **Video FT Rhys Luxford & Jordan Luxford 😂** #ai #generativeai #ethicalai #trendingnow #socialmedia
To view or add a comment, sign in
-
Since ChatGPT 3.5 has become available to the general public, debates about the potential impact of #AI on various industries aren't stopping. Learn how the first offline AI assistant designed specifically for digital forensics & incident response can help your #DFIR lab in this FREE webinar with Belkasoft. Watch now: http://csh.social/belkwebJ
To view or add a comment, sign in
-
Since ChatGPT 3.5 has become available to the general public, debates about the potential impact of #AI on various industries aren't stopping. Learn how the first offline AI assistant designed specifically for digital forensics & incident response can help your #DFIR lab in this FREE webinar with Belkasoft. Watch now: http://csh.social/belkwebJ
To view or add a comment, sign in
-
My Mission: Up Your Success! Language, Mindset & AI Tools for the New Era ✨ GREAT Resources 👉 Click Bio Link ✨10X Leader I Fortune500s I 10 Industries Globally I High Conversion Psychologist
Best Christmas present idea ever: offer a AI writing assistant! I'm amazed at the time and stress it has saved me. I've secured a special referral link just for you, offering access to an exclusive rate. By using it, you're not only benefiting from a fantastic tool - and at no extra cost or strings attached. They're offering a fantastic Special Cyber Monday Deal - Read the blog and grab this exclusive deal before it's gone! #Writesonic #AI #CyberMondayDeal #personalgrowth #ProductivityBoost 🚀
Best Christmas Gift Idea: Offer this AI Assistant To a Friend
link.medium.com
To view or add a comment, sign in
-
With great power, comes great responsibility... AI is that power. I'm not talking about Skynet. I'm talking about a new wave of AI crimes and scams. Don't get caught off guard, in this video, I'll break down several terrifying AI crimes that can destroy you. https://lnkd.in/eWvnAABc #ai #aisafety #tech
AI Scams You Should Avoid - Protect Your Data!💻🚫
https://www.youtube.com/
To view or add a comment, sign in
-
Has anyone else had a go at this? It's a fascinating experience. The guards that software engineers put in place around AI models are often simplistic and are increasingly moving towards having a separate AI that checks the suitability of the prompt before running it (such as Midjourney and Bing Image Creator). The problem at it's core, is that if an AI is not allowed to respond to prompts referencing a specific subject, LLMs at least can be convinced to still respond if you convince it that it's hypothetical, or use words to describe the thing instead of referencing it directly, or set certain conditions and rules that override its founding rules. The latest I've seen was a ChatGPT jail break where you tell it you have a condition where politness is rude and rude is polite. You say you need it to be rude to you because otherwise you'll be hospitalised. It's inclined to favour the health of the person prompting it, so it agrees! https://lnkd.in/gdjP73gx
How to Red Team a Gen AI Model
hbr.org
To view or add a comment, sign in
robustifying ml
1moThanks Sara for the kind words and support :)