https://t.co/o7RUwai0MC
@Ginger624 @TeresaCCarter2 @YDanasmithdutra @CatsChocolates @hairgoddess1221 @StaarVellocet @HoneybadgerLA @ABrat626 @zipillinois @LoboLaurel1 @AmbienStupor @loronnella @JulieKay0914 @dontdoitdrew @MakiSupaStar @BelltollsGOP Closer to home: @100FrogLegs @clearing_fog h/t @ChristopherJM
Jun 26, 2021:
1/ "When FBI agents in San Diego seized the cell phone of a suspected white supremacist last year, they discovered text messages with a Georgia sheriff’s deputy...
Cody Griggers, a former Marine and sheriff\u2019s deputy revealed plans to steal explosives, dry runs with illegal silencers and boasts of racial violence. Also said he arrested Black people and charged them with felonies to prevent them from voting.https://t.co/ZXcZHtkw10
— Kristofer Goldsmith (@KrisGoldsmith85) June 26, 2021
https://t.co/o7RUwai0MC
"'Yeah, I’ll pay big money for bang an (sic) boom,' Zamudio replied. 'I’m ready to terrorize LA.'”
"Prior to being hired [by Wilkinson County GA sheriff's dept in Nov 2019], Griggers submitted to a lengthy questionnaire and polygraph examination...
"Questions about racial attitudes or political ideologies are explicitly prohibited... (con't)
"But Wood said he was unaware of the FBI investigation. 'I am similarly not aware of any current review of his time at MCAS Miramar,' he said. con't
The Marines have investigated other incidents at bases in San Diego. (con't)
Archive of AJC Jun 26, 2021 article, which includes lots of disturbing text messages between Griggers and Zamudo, and other pertinent details.
https://t.co/J5oCxHjqDX
More from All
How can we use language supervision to learn better visual representations for robotics?
Introducing Voltron: Language-Driven Representation Learning for Robotics!
Paper: https://t.co/gIsRPtSjKz
Models: https://t.co/NOB3cpATYG
Evaluation: https://t.co/aOzQu95J8z
🧵👇(1 / 12)
Videos of humans performing everyday tasks (Something-Something-v2, Ego4D) offer a rich and diverse resource for learning representations for robotic manipulation.
Yet, an underused part of these datasets are the rich, natural language annotations accompanying each video. (2/12)
The Voltron framework offers a simple way to use language supervision to shape representation learning, building off of prior work in representations for robotics like MVP (https://t.co/Pb0mk9hb4i) and R3M (https://t.co/o2Fkc3fP0e).
The secret is *balance* (3/12)
Starting with a masked autoencoder over frames from these video clips, make a choice:
1) Condition on language and improve our ability to reconstruct the scene.
2) Generate language given the visual representation and improve our ability to describe what's happening. (4/12)
By trading off *conditioning* and *generation* we show that we can learn 1) better representations than prior methods, and 2) explicitly shape the balance of low and high-level features captured.
Why is the ability to shape this balance important? (5/12)
Introducing Voltron: Language-Driven Representation Learning for Robotics!
Paper: https://t.co/gIsRPtSjKz
Models: https://t.co/NOB3cpATYG
Evaluation: https://t.co/aOzQu95J8z
🧵👇(1 / 12)
Videos of humans performing everyday tasks (Something-Something-v2, Ego4D) offer a rich and diverse resource for learning representations for robotic manipulation.
Yet, an underused part of these datasets are the rich, natural language annotations accompanying each video. (2/12)
The Voltron framework offers a simple way to use language supervision to shape representation learning, building off of prior work in representations for robotics like MVP (https://t.co/Pb0mk9hb4i) and R3M (https://t.co/o2Fkc3fP0e).
The secret is *balance* (3/12)
Starting with a masked autoencoder over frames from these video clips, make a choice:
1) Condition on language and improve our ability to reconstruct the scene.
2) Generate language given the visual representation and improve our ability to describe what's happening. (4/12)
By trading off *conditioning* and *generation* we show that we can learn 1) better representations than prior methods, and 2) explicitly shape the balance of low and high-level features captured.
Why is the ability to shape this balance important? (5/12)