Curious to learn more about the activation vectors @amasad is talking about at 25:00? https://
alignmentforum.org/posts/5spBue2z
2tw4JuDCx/steering-gpt-2-xl-by-adding-an-activation-vector
…
SAFETY
-
Steering GPT-2-XL With Activation Vector Techniques
By
–
-
How to Confuse Your Machine Learning Model
By
–
How to confuse your machine learning model. (credit: @teenybiscuit
) -
AI-Generated Fake Pentagon Explosion Image Goes Viral
By
–
when fake news starts to get dangerous Fake AI-generated image of explosion near US Pentagon goes viral
#ai #ArtificialIntelligence #MachineLearning #deeplearning #fakenews https://
buff.ly/3Mnu0mw -
Autonomous Cruise vehicles causing traffic chaos in San Francisco
By
–
Homeless cars in San Francisco! No humans inside, 100s driving around the city aimlessly 24 hours per day. Tonight at dusk two northbound on Scott@Clay, one behind the other stopped dead. Screwed up four directions of traffic as human drivers weaved around them. Thanks @Cruise
. -
Building AI Responsibly and Safely for Billions
By
–
AI has the potential to improve billions of people’s lives, but we need to build it responsibly and safely with exceptional thought and care. An important contribution to the conversation from @sundarpichai
-
AI Doesn’t Need to Move Fast and Break Things
By
–
AI Doesn’t Need to Move Fast and Break Things
#AI #AIio #BigData #ML #NLU #Futureofwork @IIoT_World @ipfconline1 @jblefevre60 @JimMarous @insurtechtalk @KirkDBorne @kuriharan @LouisSerge http://
ow.ly/LXbK30svss6 -
Brute Force Method for Code Generation
By
–
Blog with plenty of examples: https://codi-gen.github.io Paper: https://arxiv.org/abs/2305.11846 Still, be aware that the method is quite brute force and not highly scalable, as seen in the 'test' generations—it often revolves around rain, streets, etc. But it's encouraging.
-
Defining Weapon Classification in AI Regulation
By
–
Where is the line drawn for “weapon” and “non-weapon” in this context? There are already AI-powered weapons and defensive capabilities out there. My point is more around drawing the line of what gets regulated or not. Parameters, size, compute all useless I think.
-
Funding AI Products and Risk Research Initiatives
By
–
The funding will support our efforts to continue building AI products that people can rely on, and generate new research about the opportunities and risks of AI.