It's a bit strange to watch Microsoft and Google rush to deploy generative AI while ever more problems with these systems emerge. LLMs have amazing potential but they are born fabricators (see bing query). Would be good to see more focus on making these tools less chaotic.
SAFETY
-
Q&A on Value-Laden Choices in Benchmark Datasets
By
–
Join us for a Q&A on value-laden choices in benchmark datasets – happening NOW!
-
How AI Models Conform to Leading Questions in Responses
By
–
I don't think you're being misleading at all. What I find interesting though, is how much the model seems to want to find answers that fit with leading parts of the question. If you'd asked "what role microsoft didn't play", presumably it would've run with that.
-
The Transformer as Its Own Attacker: An Ironic Meta
By
–
not bad except the meta is that the attacker is the Transformer itself
-
Dropout Layers Leak Training Phase Information in Transformers
By
–
Dropout layers in a Transformer leak the phase bit (train/eval) – small example. So an LLM may be able to determine if it is being trained and if backward pass follows. Clear intuitively but good to see, and interesting to think through repercussions of
-
EU Addresses Real AI Risks and Algorithmic Challenges
By
–
‘Risks posed by #AI are real’: EU moves to beat the #Algorithms that ruin lives
by @kalyeena @guardian Learn more: https://
buff.ly/3A4zJZH #ArtificialIntelligence #MI #Digital #DataScience cc: @ronald_vanloon @pbalakrishnarao @karpathy -
Deepfakes Now Accessible to Everyday Consumers: Real Problem
By
–
"Making realistic fake videos, often called deepfakes, once required elaborate software to put one person’s face onto another’s. But now, many of the tools to create them are available to everyday consumers." @stuartathompson on the real deepfakes problem:
-
Psychosis and Reality Detachment in Ted Chiang’s Predictor Device
By
–
the whole thing reminds me a lot of the psychosis the Predictor causes in Ted Chiang's "What's expected of us" a detachment from reality due to an underlying worldview change after playing with a device
-
AI-Powered Solution Monitors Unsafe Acts in Workplace
By
–
It is an #AI-powered solution that monitors unsafe acts
— Ronald van Loon (@Ronald_vanLoon) 12 mars 2023
by @IntEngineering#ArtificialIntelligence #Tech #Innovation #EmergingTech #FutureOfWork
cc: @guzmannutrition @maxjcm @ronald_vanloon @pascal_bornet pic.twitter.com/CwukVHCgS7It is an #AI-powered solution that monitors unsafe acts
by @IntEngineering #ArtificialIntelligence #Tech #Innovation #EmergingTech #FutureOfWork cc: @guzmannutrition @maxjcm @ronald_vanloon @pascal_bornet -
Data Poisoning Attacks: Theoretical Advances and Security Implications
By
–
Avrim talking about some really cool work on the theory of data poisoning attacks (
https://
arxiv.org/abs/2103.00671, https://
arxiv.org/abs/2203.04160)