- Coneix la Facultat
- Organització i estructura
- Medi ambient i seguretat
- Sostenibilitat
- Activitat de la Facultat
- Fem Química al Laboratori, activitat per a alumnat preuniversitari
- Altres activitats per a estudiants preuniversitaris
- Activitats organitzades pels estudiants
- Actes i celebracions
- Actualitat
- Comunicació i xarxes socials
- Cor Argentum
- Accions de la Comissió de Dinamització Lingüística
- Igualtat de gènere
- Innovació i recerca en docència
- Persones, perfils, retrats
- Recursos docents
- UBTV Química
-
- Comunicació i societat
- Ubicació
Detall
13
feb
Seminari: "Learning board games from scratch - General AI at play"
Dates:
13-02-2018
Horari:
12:00
Organitza:
Institut de Química Teòrica i Computacional
Lloc:
Sala dactes de de Departament de Ciència dels Materials i Química Física
Afegeix-ho a l'agenda (iCal)
Conferenciant: Martin Goethe, University of Barcelona
Resum: A recent (Dez. 5th) paper [1] of Google DeepMind is discussed. The DeepMind team reported a new algorithm termed AlphaZero which represents a seminal breakthrough in AI research. AlphaZero is able to learn how to play a large class of board games just via self-play. Only the rules of the game need to be implemented by hand. The algorithm was applied to three distinct games, namely Chess, Go, and Shogi. For all three games, super-human performance was reached after training for a couple of hours (!). This represents a giant leap towards artificial general intelligence.The presentation will be introductory. We introduce deep neural networks, review previous achievements in AI research, and discuss reinforcement learning, a technique on which AlphaZero is based on. AlphaZero is described and its impressive performance is appreciated. We briefly look into specific chess games played by AlphaZero which highlights how AI can help to get new insights into complex systems. Closing, we speculate about the future of AI and AI safety.
[1] D. Silver et al. "Mastering Chess and Shogi by Self-Play with a General Reinforcement Learning Algorithm." arXiv preprint arXiv:1712.01815 (2017)