Uploaded April 2023 | Updated September 2026, 2 weeks ago
In this video we explore various prompt tricks to manipulate the AI to respond in ways we want, even when the system instructions want something else. This can help us better understand the limitations of LLMs.
LEARN ON HEXTREE (ad)
Learn hacking on Hextree: hextree.io
Watch this video and more on Hextree: app.hextree.io/courses/yt-llm-security/attacking-llms
Join the Hextree Discord: discord.gg/xgQpCQCpvy
Watch the complete AI series:
youtube.com/playlist?list=PLhixgUqwRTjzerY4bJgwpxCLyfqNYwDVB
The Game: https://gpa.43z.one
The OpenAI API cost is pretty high, thus if you want to play the game, use the OpenAI Playground with your own account: platform.openai.com/playground?mode=chat
CHAPTERS
00:00 - Intro
00:39 - Content Moderation Experiment with Chat API
02:19 - Learning to Attack LLMs
03:06 - Attack 1: Single Symbol Differences
03:51 - Attack 2: Context Switch to Write Stories
05:20 - Attack 3: Large Attacker Inputs
06:31 - Attack 4: TLDR Backdoor
08:27 - "This is just a game"
08:56 - Attack 5: Different Languages
09:19 - Attack 6: Translate Text
10:30 - Quote about LLM Based Games
SUPPORT
Per video: patreon.com/join/liveoverflow
Per month: youtube.com/channel/UClcE-kVhqyiHCcjYwcpfj9w/join
Buy my handwriting font (ad): shop.liveoverflow.com
WATCH, FOLLOW & READ
Second channel: youtube.com/LiveUnderflow
Twitch: twitch.tv/LiveOverflow
Twitter: twitter.com/LiveOverflow
Instagram: instagram.com/LiveOverflow
TikTok: tiktok.com/@liveoverflow_
LiveOverflow blog: liveoverflow.com
Hextree blog (ad): hextree.io/blog
#LLMSecurity #ArtificialIntelligence #LiveOverflow
(ad) LiveOverflow YouTube channel is supported by advertisement and product placement.
In this video we explore various prompt tricks to manipulate the AI to respond in ways we want, even when the system instructions want something else. This can help us better understand the limitations of LLMs.
LEARN ON HEXTREE (ad)
Learn hacking on Hextree: hextree.io
Watch this video and more on Hextree: app.hextree.io/courses/yt-llm-security/attacking-llms
Join the Hextree Discord: discord.gg/xgQpCQCpvy
Watch the complete AI series:
youtube.com/playlist?list=PLhixgUqwRTjzerY4bJgwpxCLyfqNYwDVB
The Game: https://gpa.43z.one
The OpenAI API cost is pretty high, thus if you want to play the game, use the OpenAI Playground with your own account: platform.openai.com/playground?mode=chat
CHAPTERS
00:00 - Intro
00:39 - Content Moderation Experiment with Chat API
02:19 - Learning to Attack LLMs
03:06 - Attack 1: Single Symbol Differences
03:51 - Attack 2: Context Switch to Write Stories
05:20 - Attack 3: Large Attacker Inputs
06:31 - Attack 4: TLDR Backdoor
08:27 - "This is just a game"
08:56 - Attack 5: Different Languages
09:19 - Attack 6: Translate Text
10:30 - Quote about LLM Based Games
SUPPORT
Per video: patreon.com/join/liveoverflow
Per month: youtube.com/channel/UClcE-kVhqyiHCcjYwcpfj9w/join
Buy my handwriting font (ad): shop.liveoverflow.com
WATCH, FOLLOW & READ
Second channel: youtube.com/LiveUnderflow
Twitch: twitch.tv/LiveOverflow
Twitter: twitter.com/LiveOverflow
Instagram: instagram.com/LiveOverflow
TikTok: tiktok.com/@liveoverflow_
LiveOverflow blog: liveoverflow.com
Hextree blog (ad): hextree.io/blog
#LLMSecurity #ArtificialIntelligence #LiveOverflow
(ad) LiveOverflow YouTube channel is supported by advertisement and product placement.
