Moderne KI-Agenten handeln – immer auf maximale Belohnung aus – intuitiv und rational – 📘 Neuer Artikel von Dr. Matthias Unverzagt
#ittage #KI #ReinforcementLearning #chatgpt
https://www.informatik-aktuell.de/betrieb/kuenstliche-intelligenz/moderne-ki-agenten-handeln-intuitiv-und-rational.html

Evaluating choices: computational analysis of 7 studies suggests that internally defined goals play a crucial role in shaping the subjective value attributed to available options in #ReinforcementLearning @gaia_molinaro @Anne_On_Tw @ccnlab #PLOSBiology https://plos.io/3pPSJsH

Today we’re diving into coding #ReinforcementLearning for our fisheries examples.

#AMSI winter school at #QUT

Carl Boettiger has us playing in GitHub Codespaces trying to maximise the number of fish we harvest. We’ll be learning about #ReinforcementLearning this week

#mathematics #modelling #modeling #AMSIWinterSchool #QUT

I've uploaded our BBS paper on "proxy failure" — analogues of Goodhart's Law across scales ranging from molecular biology to neuroscience to business to ecology — on ResearchGate.

https://www.researchgate.net/publication/371866602_Dead_rats_dopamine_performance_metrics_and_peacock_tails_proxy_failure_is_an_inherent_risk_in_goal-oriented_systems

#Neuroscience #Biology #Economics #Ecology #Evolution #ReinforcementLearning

What is the #Universe expanding into? : Medium

Scientists Discover a #VirginBirth in a #Crocodile : NY Times

Faster #Sorting #Algorithms discovered using deep #ReinforcementLearning : Nature

Check our latest #KnowledgeLinks

https://knowledgezone.co.in/resources/bookmarks

Interesting interview which mentions #ReinforcementLearning with human feedback #rlhf.

“#ChatGPT architect John Schulman discusses his journey with #AI during UC Berkeley visit”

https://flip.it/QFVd0g

#AIshift #NewPhaseOfAI #RuleBasedAlgorithms #MachineLearning #DeepLearning #ReinforcementLearning #NaturalLanguageProcessing #EthicalAI #TransparentAI #ExplainableAI

InstructRL, or how to leverage an #LLM to tune an #ReinforcementLearning based agent: https://arxiv.org/pdf/2304.07297.pdf

The #LLM combines the observed state with a natural language instruction and produces a policy (the tasks matching the instruction). This policy is not right, but serves as a prior for the RL-based model, which will produce a good policy close to the prior.

It results in a practical way to declare user preferences to a well controlled RL-based #AI agent.

(1/2) MIT launched the 2023 edition of the Introduction to Deep Learning course 🚀 by Alexander Amini and Ava Amini. The course started in March and will run until May. All the course materials are available, and it covers the following topics:
✅ Deep learning foundation
✅ Computer vision
✅ Deep generative modeling
✅ Reinforcement learning
✅ Robot learning
✅ Text to image

#deeplearning #datascience #machinelearning #reinforcementlearning #computervision #python #tensorflow

(1/2) Stanford University released yesterday a new course - Deep Multi-Task and Meta-Learning 🚀. Prof. Chelsea Finn teaches the course, and it focuses on the following:
✅ Foundations of modern deep learning methods for learning across tasks
✅ Implement and work with practical multi-task & transfer learning systems (in PyTorch)
✅ A glimpse into the scientific and engineering process of building and understanding new algorithms
#deeplearning #MachineLearning #reinforcementlearning #DataScience

I am looking forward to test #Claude #LLM, one of the main competitors to #ChatGPT. It is different because it is better designed to respect constraints to make it harmless, thanks to #ReinforcementLearning : https://arxiv.org/pdf/2212.08073.pdf

But there is a trade-off between usefulness and harmlessness, and it is effectively assessed in the paper!

#IBM going at #LLMs for #Watson #AI, with a bit of their touch: https://arxiv.org/pdf/2303.05510.pdf

They improve the output by guiding the #transformer with #ReinforcementLearning-based planning. It is used to look ahead and check whether the transformer’s output is valid.

The #CIFAR #deeplearning and #reinforcementlearning Summer School is open for applications (deadline Feb 22):

https://dlrl.ca/

This is a chance to learn AI from some of the top-minds in the field, here at Mila! Don't miss it!

I loved Hugo Casselles-Dupré’s presentation in #TalkingRobotics:
https://www.youtube.com/watch?v=i4ovDv8DdzE

He makes a #robot learn better from another robot’s demonstration by leveraging #pragmatics! #machinelearning #ReinforcementLearning

Using #Bayesian inference, the teacher would deduce the goal behind its candidate demonstration and check that it’s not ambiguous for the learner. The learner also learns faster by guessing the intended goal of the teacher.

RT @ak_poorni
If you are interested in #ComputationalPsychiatry #ComputationalModeling #MachineLearning #ReinforcementLearning and looking for a postdoctoral position, apply to become a T32 trainee @McLeanHospital with me :)

I love the work of MLU-Explain - part of Amazon Machine Learning University, and their brilliant usage of data visualization to explain data science concepts ❤️. Last week they released a new article explaining Reinforcement Learning 👇🏼
https://mlu-explain.github.io/reinforcement-learning/

I highly recommend checking other articles available on the MLU-Explain:
https://mlu-explain.github.io/

Great work Jared Wilber, Erin Bugbee, and Anand Kamat! 🙏🏼

#DataScience #MachineLearning #dataviz #infographic #reinforcementlearning

Deep Reinforcement Learning Course Notes 🚀🚀🚀

Holy 🐮, it is amazing to see the level of detail in Hadar Shavit lecture's notes 🤯. If you are looking for a short summary of Deep Reinforcement Learning, I recommend checking Hadar's notes he took while taking the Deep Reinforcement Learning course at the Ben-Gurion University of the Negev 👇🏼

https://github.com/Hadar933/Deep-Reinforcement-Learning

Thanks to Hadar for creating this doc and sharing it with others! 🙏

#deeplearning #reinforcementlearning #machinelearning

RT @ak_poorni@twitter.com

Computational Psychopathology (COMP) group at McLean Hospital/Harvard Medical School (Directed by me) is looking for a postdoc interested in #ComputationalPsychiatry to work on developing computational models of #ReinforcementLearning and #Decisionmaking. Please RT.

🐦🔗: https://twitter.com/ak_poorni/status/1605281321106149376

👀🚨Exciting post-doc opportunity 🚨👀 in #ComputationalPsychiatry & #ReinforcementLearning working with Dr. Poornima Kumar!
I'm really excited to see what comes out of this group!!!
---
RT @ak_poorni
Computational Psychopathology (COMP) group at McLean Hospital/Harvard Medical School (Directed by me) is looking for a postdoc interested in #ComputationalPsychiatry to work on developing computational models of #ReinforcementLearning and #Dec…
https://twitter.com/ak_poorni/status/1605281321106149376