Скачать или смотреть ☠️🤖 Data Poisoning and AI Trust: A Growing Security Challenge

What if AI systems we rely on every day were secretly trained on manipulated data?

Let’s chat about data poisoning - a security challenge that can have wide effects, from search results to drug discovery, and what we can do about it.

🎯 Why this matters for everyday AI use
🧠 How misinformation gets "frozen" in AI systems
🛡️ Steps we can take to stay safer

P.s. Recorded in November.

📚 Resources for Deeper Learning:
Understanding the Challenge
"Poisoning Web-Scale Training Datasets is Practical" (Research Paper)
https://arxiv.org/abs/2302.10149v2
"Forcing Generative Models to Degenerate Ones" (Research Paper)
https://arxiv.org/abs/2312.04748v1

SEO & Information Quality
"The People Who Ruined the Internet" (The Verge)
https://www.theverge.com/features/239...

AI Literacy & Education
"Make AI Literacy a Priority With These Free Resources"
https://www.gettingsmart.com/2024/06/...

💭 What worries you more - AI making honest mistakes, or someone deliberately misleading it? Share your thoughts! I'm really curious about your experience with AI reliability and how you handle it.

Chapters
0:00 - The Growing Challenge 🌟
1:00 - Misinformation Networks 🥸
1:30 - Training Data Attacks 🏴‍☠️
2:27 - Current Solutions 🛠️
2:49 - SEO on Google 🔎
3:19 - What to do about it 📚
4:53 - Conclusion 🤔