"Watermarking Language Models" paper and GPTZero EXPLAINED | How to detect text by ChatGPT?

Описание: Did ChatGPT write this text? We explain two ways of knowing whether AI has written a text: GPTZero and Watermarking language models.
► Sponsor: Cohere 👉 https://t1p.de/22srn

UPDATE: Further research about the Reliability of Watermarks for LLMs! https://arxiv.org/abs/2306.04634 shows that watermarking works even when watermarked text is re-written by humans or paraphrased by another non-watermarked LLM.

Check out our daily #MachineLearning Quiz Questions:    / aicoffeebreak  
➡️ AI Coffee Break Merch! 🛍️ https://aicoffeebreak.creator-spring....

📜 Watermarking LMs paper: Kirchenbauer, John, Jonas Geiping, Yuxin Wen, Jonathan Katz, Ian Miers and Tom Goldstein. “A Watermark for Large Language Models.” (2023). https://arxiv.org/abs/2301.10226
📖 Tom Goldstein’s tweet summary:   / 1618287665006403585  
🔗 GPTZero: https://gptzero.me/
🔗 First version of GPTZero (easy and free to try out): https://etedward-gptzero-main-zqgfwb....

Thanks to our Patrons who support us in Tier 2, 3, 4: 🙏
Dres. Trost GbR, Siltax, Edvard Grødem, Vignesh Valliappan, Mutual Information, Mike Ton

Outline:
00:00 Did ChatGPT generate this text?
01:14 Cohere [Sponsor]
02:43 How does GPTZero work: Perplexity explained
05:42 GPTZero “Burstiness”
07:08 Robust detection with watermarking
07:46 Language modelling decoding explained
09:21 Watermarking explained
11:05 “soft” watermark
12:49 Attacking watermarks

▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀
🔥 Optionally, pay us a coffee to help with our Coffee Bean production! ☕
Patreon:   / aicoffeebreak  
Ko-fi: https://ko-fi.com/aicoffeebreak
▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀

🔗 Links:
AICoffeeBreakQuiz:    / aicoffeebreak  
Twitter:   / aicoffeebreak  
Reddit:   / aicoffeebreak  
YouTube:    / aicoffeebreak  

#AICoffeeBreak #MsCoffeeBean #MachineLearning #AI #research​
Music 🎵 : Sunset n Beachz - Ofshane
Video editing: Nils Trost