[#Article] Japan: Ministry of Defense presents its 1st basic policy on AI use https://actuia.com/actualite/japon-le-ministere-de-la-defense-presente-sa-1ere-politique-de-base-sur-lutilisation-de-lia/
… #AI #ArtificialIntelligence
SAFETY
-
Japan: First Defense Policy on AI
By
–
-
Responsible Computing for AI Safety and Trust in India
By
–
Meet the speaker @ #GlobalIndiaAISummit -Mr. Jibu Elias, Country Lead for the Responsible Computing Challenge, Mozilla Foundation. Watch as he shares his insights on how to accomplish the core pillar of safety and trust in India's AI mission. Catch him live at the… pic.twitter.com/2arppmMHuO
— IndiaAI (@OfficialINDIAai) 10 juillet 2024Meet the speaker @ #GlobalIndiaAISummit -Mr. Jibu Elias, Country Lead for the Responsible Computing Challenge, Mozilla Foundation. Watch as he shares his insights on how to accomplish the core pillar of safety and trust in India's AI mission. Catch him live at the
-
Confidence in AI beliefs parallels religious certainty and uncertainty
By
–
As I read these posts I find it intriguing to see how confident people are about their beliefs; on both sides, while the future fate of AI seems truly very uncertain to me. I see parallels with religion (my God the only real one). At least one must have badly calibrated beliefs.
-
Data Poisoning Removal: Hypotheses on Model Parameter Impact
By
–
We present a few hypotheses as to why data poisoning is so tricky to remove. Some hypotheses include that data poisoning moves the model parameters a lot, or in a subspace orthogonal to the clean data. Would love to see more exploration here! 6/n
-
Data Poisoning Attacks Remain Effective Against Unlearning Methods
By
–
We investigate the effect of: three types of data poisoning attacks (targeted, indiscriminate, and "Gaussian," which is new), across CNNs and LLMs, for seven different unlearning methods. None of these are effective at removing the effect of poisoning. 5/n
-
MUL Method Mitigates Data Poisoning Effects in Training
By
–
But MUL has also been proposed as a method to remove the effect of data poisoning. Such effects are *indirect* — they manifest in other datapoints besides the specific training point. 4/n
-
Evaluating Machine Unlearning with Membership Inference Attacks
By
–
The most common way to evaluate an MUL method is using "membership inference attacks." These try to directly test whether the "unlearned" point was in the dataset used to train the model or not. 3/n
-
Machine Unlearning Fails Against Data Poisoning Attacks
By
–
New paper: Machine Unlearning Fails to Remove Data Poisoning Attacks, ft @MartinPawelczyk
, @jimmy_di98
, @ayush_sekhari
, @SethInternet
. Title says it all: current approaches for machine unlearning (MUL) are not effective at removing the effect of data poisoning attacks. 1/n -
Koala Lifter: Wind Turbine Self-Climbing Robot System
By
–
Meet the Koala Lifter 🐨!
— Pascal Bornet (@pascal_bornet) 9 juillet 2024
A disruptive self-climbing system that uses the strength of the wind turbine tower as support to climb up the turbine.
No more cranes are needed!#Automation #Robotics #Safety #Innovation pic.twitter.com/sMdo4WcSujMeet the Koala Lifter ! A disruptive self-climbing system that uses the strength of the wind turbine tower as support to climb up the turbine. No more cranes are needed! #Automation #Robotics #Safety #Innovation
