While I'm here, I also want to note that @DAIRInstitute
, in addition to uncovering + mitigating issues w.r.t. AI, we "cultivate spaces to accelerate imagination & creation of new technologies & tools to build a better future." https://
dair-institute.org/research-philo
sophy/
… has our research philosophy.
SAFETY
-
DAIR Institute: Mitigating AI Issues and Creating Future Technologies
By
–
-
AI and Web3 ID Technology Combat Deepfake Fraud
By
–
How AI And Web3 ID Tech Can Defeat Deepfake Frauds
#AI #AIio #BigData #ML #NLU #Futureofwork @CRudinschi @AntonioSelas @alexjc @RobotLaunch @karpathy @andyjankowski @bobgourley @CadeMetz http://
ow.ly/3qI230swABx -
AI Safety Discourse and the Doomer Debate in Research
By
–
Thank you to Lorena for writing this article, highlighting our works and showing that this is a collective effort. I like that you didn't equivocate about the fact that I was fired. The part about Hinton was such a great example of where we are with the doomers
-
Deep Learning Limitations Unchanged Since 2017
By
–
Re-reading an article I wrote in 2017, and I'm finding I could have written it yesterday. It's fascinating how the limitations of deep learning have stayed the same since. Same problems, same failure modes, no progress.
-
Media bias: forced positivity narratives despite documented AI harms
By
–
It's part and parcel of that annoying trope of having to pay lip-service to the supposed up-sides of "AI" even in reporting about the actual (documented, really happening) harms & risks of further harm. >>
-
System Safety Boundaries and Reasonable Use Cases
By
–
To my knowledge, no one was hurt. But I think this case is instructive for what might happen in similar situations when the boundaries of reasonable use of the system are less obvious.
-
Synthetic Media Systems Must Handle Unpredictable User Inputs
By
–
Today's case study in: If you're going to set up a synthetic media machine where anyone can access it, you'll need to account for people throwing all kinds of random stuff at it. >>
-
AI-Generated Videos Will Soon Flood Our Social Media Feeds
By
–
Dang, we gonna have believable videos like this saturating our feeds sooner than I expected
-
LLMs Unreliability Demands Regulatory Framework
By
–
LLMs are wildly unreliable; that’s part of why they need to be regulated.
-
Waymo Safety Report: Analysis of Autonomous Vehicle Incidents
By
–
The technical report from the page above is worth reading: https://
storage.googleapis.com/waymo-uploads/
files/documents/safety/SafetyPerformanceofWaymoROat1Mmiles.pdf
… Pages 11 to 13 have a detailed description of each of the 20 incidents (spoiler: "Human drivers violated road rules and/or behaved dangerously in every vehicle-to-vehicle event").