Montréal.IA(@Montreal_IA) 's Twitter Profile Photo

Deeds carry more weight than words . . .

& lay emphasis on actions, while is confined to predicting the subsequent word.

Corollary : will supersede ChatGPT

∴ Henceforth: AGI Agent  > ChatGPT

Agent

Deeds carry more weight than words . . .

#MetaLearning & #ReinforcementLearning lay emphasis on actions, while #ChatGPT is confined to predicting the subsequent word.

Corollary : #RL will supersede ChatGPT

∴ Henceforth: AGI Agent  > ChatGPT

#AGI #AGIAgent #MontrealAI
account_circle
Dimitri Bertsekas(@DBertsekas) 's Twitter Profile Photo

My course at ASU has been completed. Slides, videolectures, and a new 400-page textbook “A Course in Reinforcement Learning”, to be published soon, can be found at web.mit.edu/dimitrib/www/R…

account_circle
The Learning Guild #DevLearn #LSCon #Learning23(@LearningGuild) 's Twitter Profile Photo

Calling all learning enthusiasts! Don't miss our free webinar: Lessons from Classic Games to Design Games for Learning. Join us on June 21 and uncover the history of learning games and learn how to incorporate their insights into your own designs.

account_circle
Ryoma Sato(@joisino_en) 's Twitter Profile Photo

Need data for machine learning? We propose Seafaring, a method for automatically acquiring useful data from the Web by regarding the myriad data on the Web as a huge pool of active learning.

Paper📜: arxiv.org/abs/2210.08205
Code📂: github.com/joisino/seafar…

Need data for machine learning? We propose Seafaring, a method for automatically acquiring useful data from the Web by regarding the myriad data on the Web as a huge pool of active learning.

Paper📜: arxiv.org/abs/2210.08205
Code📂: github.com/joisino/seafar…
account_circle
T.Yasaki(@ZappyZappy7) 's Twitter Profile Photo

マルチタスク強化学習フレームワークの訓練により、多種多様なジャンプをこなす二足歩行ロボット
youtu.be/aAPSZ2QFB-E

account_circle
Tianyi Zhou(@zhoutianyi) 's Twitter Profile Photo

Continual on new tasks is a natural human skill but how to train a meta-policy that maximally reuses previous tasks' knowledge (stability) and meanwhile quickly adapts to new tasks (plasticity) ? Our recent work CoTASP addresses it by sparse prompting:

Continual  #Reinforcementlearning on new tasks is a natural human skill but how to train a meta-policy that maximally reuses previous tasks'  knowledge (stability) and meanwhile quickly adapts to new tasks (plasticity) ? Our recent work  CoTASP addresses it by sparse prompting:
account_circle
Covariant(@CovariantAI) 's Twitter Profile Photo

At just 21, Rocky Duan became one of OpenAI's early hires, pushing boundaries on , and imitation learning.

At 23, he went on co-found Covariant bringing breakthrough ML concepts to robotic automation for , where he continues to serve as CTO.

account_circle
Fordham Institute(@educationgadfly) 's Twitter Profile Photo

Advanced education is critical for student growth. From acceleration to equitable achievement grouping to selective enrollment schools, we must help students reach their full potential. Read 'Building a Wider, More Diverse Pipeline of Advanced Learners.' fordhaminstitute.org/national/resea…

account_circle