Giovanni Franzese

Lead Researcher on Humanoid Manipulation @ Technology Innovation Institute.

PhD Researcher at @tudelft.

Alumnus @polimi, @tueindhoven and @ucl.

[Chubby♨️ @kimmonismus]

Okay, that’s crazy. OpenAI actually paused its $200 subscription plan because GPT-6 Astra is such a huge success.

What a year for OpenAI

[Giovanni Franzese @franzesegi]

@visegrad24 @elmanuel2992

[Giovanni Franzese @franzesegi]

@miniapeur So good meme! So many times I had to explain that what I write in paper is not supposed to be read as math but as pseudocode...

[Physics Memes @ThePhysicsMemes]

someone at grammarly didn't pass classical mech #ITeachPhysics

[The Struggling Scientists @TheStrugglingS4]

Hello darkness my old friend🥲

[Giovanni Franzese @franzesegi]

If you want to learn how to use Human Feedback in Robotics, this is a recent book that I recently published with my colleagues at TU Delft!
You would be surprised on discovering how efficient and effective #humanfeedback is!
https://t.co/uctkRLWOfs

[Nando de Freitas @NandoDF]

People are asking if there are alternatives to RL in RLHF. Yes, imitation with DAGGER (tutorial: https://t.co/aEMx5xDIOU ). The user provides feedback with corrections, e.g. when the agent says “that” the user tells the agent that instead of saying “that”, it should say “this”.

[Giovanni Franzese @franzesegi]

Human feedback is like the secret ingredient in a recipe for #MachineLearning and #robotics success. It's the key to creating models that truly understand and serve us! #AI #hunan #Feedback.
Look how I train a policy to swing up a pendulum in 1 min: https://t.co/kzLrUtiz3D

[Academic Chatter™ @AcademicChatter]

Surround yourself with supportive people, whether they are friends, family, or fellow graduate students. Your trusted network can provide you with emotional support, feedback, stress relief, motivation, and fun, as you work towards overcoming challenges and achieving success.

[Giovanni Franzese @franzesegi]

Robotics lesson: hard code whatever you can; learn whatever is left/too hard to model. Using machine learning for the sake of using it is not only unnecessary but stupid. Check out how classic control, simple rules and learning from human, empower robots
https://t.co/SlUreqiqYT

[TU Delft @tudelft]

“Freely accessible or not, a PDF full of messy data is useless to me as a fellow researcher. Publishing data and code openly is only of value if someone else can work with it,” Adarsh Kalikadien. How his frustration lead to inspiration for #OpenScience: https://t.co/nMp9NaUrga