The elections in Bavaria show that the German political system, which has long seemed reasonably stolid, is in enormous flux at the moment.
A thread with four lessons. (Just wait till you get to Number 4! 😉)
Anyone who tries to sell that as a success for the liberal left is kidding themselves.
That narrative is certainly confirmed today, with the SPD in single digits.
But...
This is the CSU’s worst result in history. It comes at a moment when its sister party, Angela Merkel’s CDU, at 26%, polls worse than at any point since the 1950s.
But the Greens overwhelmingly drew voters from the SPD. So this is not the rise of a new kind of politics; it’s a realignment within the center-left camp.
Historically, Social Democrats have done so well by holding together a coalition of working-class voters (e.g. steelworkers) and the liberal bourgeoisie (e.g. teacher, university students, civil servants, artists).
Working-class voters are flocking to right-wing populists. The liberal bourgeoisie feels better represented by cosmopolitan parties like the Greens.
Cosmopolitain parties that don’t have to serve working-class voters have much more traction with the liberal bourgeoisie.
Parties like the Greens are very well set up to serve the ~20% of the education that is educated, reasonably affluent, predominantly urban, and pretty cosmopolitan in values.
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)
