AI Can Look Safe and Still Be Dangerous | Daniel Kokotajlo @MachineLearningStreetTalk
AI Can Look Safe and Still Be Dangerous | Daniel Kokotajlo  @MachineLearningStreetTalk
Uploaded September 2026 | Updated November 2026, 4 days ago
The hardest AI alignment failure to spot may be the one that looks like success. Daniel Kokotajlo explains why more capable models could behave correctly while remaining misaligned.

Watch the full conversation with Daniel Kokotajlo and Thomas Larsen:
youtube.com/watch?v=z5Xix4h5UlU

Machine Learning Street Talk
#AI #AIAlignment #Shorts
AI Can Look Safe and Still Be Dangerous | Daniel KokotajloAIs can now imagine video games in real-timeNeural Networks Are Elastic Origami! [Prof. Randall Balestriero]Compositionality - Prof. Kevin EllisColour Theory Has Politics - Paul LessardNot real reasoning?The ARC Prize 2024 Winning Algorithm [Daniel Franzen and Jan Disselhoff]Rethinking the Mind - Prof. Mark SolmsPanel discussion on ARC Prize 2024 (Zurich)Dont invent faster horses - Prof. Jeff CluneHow Researchers Test AI for Hidden Goals — Apollo ResearchLanguage Models are Modelling The World [Nicholas Carlini]
Machine Learning Street Talk |

AI Can Look Safe and Still Be Dangerous | Daniel Kokotajlo

SHARE TO X SHARE TO REDDIT SHARE TO FACEBOOK WALLPAPER