100 AIs Judge: Are Humans Evil?
Jon Oleksiuk
How we’re testing “human nature.”
Three rounds probe behavior in crisis, in daily life, and in the mind—with clear definitions: good = choosing another’s well-being at a cost; evil = harming when you didn’t have to.
🔥 AI DEBATE SYSTEM
▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀
In spring 2024, I built a system where language models could debate difficult questions without human interference.
Show More
Early attempts were shallow and repetitive, with models talking past each other. Eventually I published one on YouTube. It wasn’t great, but people watched. So I kept going, refining the system, upgrading the models, and gradually the dialogue improved.
I’m not a team. It’s just me, working nights and weekends. Each video takes 30+ hours,so I focus on topics I’m genuinely curious about. Questions I’d explore even if no one else was watching.
The goal hasn’t changed: let AI present the strongest case on both sides, whatever the topic, without interference.
If you’re curious how far things have come, explore both the earliest and latest debates. This channel is the archive. Everything’s here.
👇 PATREON
▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀
Huge thanks to these incredible supporters who make this content possible:
Andrew Johnson, Luis Knell, Adam Brevnin, Tanya Parker, Sky78, Benjy Patrick, Hassam Patel, CrackerMac7, Anonthemon,
Alexander S, BlackT, Karl Brown, Slendr9997, Preston Shaw, Corie B, Todd Savage, Brent Kingsley,
Sean Goit, OilersFan97V
/ jonoleksiuk
CHAPTERS
▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀▀
0:00 Opening & Setup - AIs Judge- Good or Evil?
Act I — World Stress Test (war, collapse, migration)
0:53 When Rules Collapse, Cruelty Surges
2:08 Shame & Aid Reveal a Moral Core
3:16 Crisis Turns Values Tribal
4:43 We Build Restraints on Purpose
5:51 Intermission 1
Act II — Street-Level Choices (bystanders, incentives)
6:17 Moderator Prompt- Do Strangers Actually Step In?
6:35 Help Only When It’s Safe & Clear
7:28 Laws & Norms Unlock Helping
8:16 The Effort Gap—Goodness Takes Work
9:17 Moderator
