ALL OF EDUCATION, FOR ALL. For the purpose of achieving all of education for all, the Intelligence Foundation Charity, in collaboration with Montreal.AI, is developing a teacher, Saraswati AI, and an agent learning to orchestrate synergies, Polymatheia AI.
"(…) in full and equal opportunities for education for all (…)" - Preamble to UNESCO’s Constitution
Saraswati AI is a novel, collaborative and open humanitarian AI project developed by the people,for the people.
Expected Beneficiaries: 100 million people.
This humanitarian AI project will be developed by the people, for the people.
In order to inspire, support and train the people who will shape the 21st Century, the Intelligence Foundation is preparing to offer academic training and conferences, prizes and recognitions and open research and publications.
- Curriculum Learning
- DeepLearning
- MetaLearning
- OpenAI Gym
- Neuroevolution
- Reinforcement Learning
- Self-Attention
- Self-Supervised Learning
- Symbolic AI (MCTS)
- Transformers
- Objective function design: Directly optimize for what we want.
- Usage of functions which modify the target density to enable aggressive data augmentations.
- Searching and Learning in the knowledge space with MCTS and Deep Reinforcement Learning.
- The Consciousness Prior, Yoshua Bengio, 2019: https://arxiv.org/abs/1709.08568v2.
- Discovering Symbolic Models from Deep Learning with Inductive Biases, Cranmer et al., 2020: https://arxiv.org/abs/2006.11287.
- Designing neural networks through neuroevolution, Stanley et al., 2019: https://www.nature.com/articles/s42256-018-0006-z.pdf.
- AI-GAs: AI-generating algorithms, an alternate paradigm for producing general artificial intelligence, Jeff Clune, 2019: https://arxiv.org/abs/1905.10985.
- Predicting What You Already Know Helps: Provable Self-Supervised Learning, Lee et al., 2020: https://arxiv.org/abs/2008.01064.
- Illuminating search spaces by mapping elites, Mouret, J.-B. and Clune, J., 2015: https://arxiv.org/abs/1504.04909.
- Generative Language Modeling for Automated Theorem Proving, Stanislas Polu and Ilya Sutskever, 2020: https://arxiv.org/abs/2009.03393.
- Learning to summarize from human feedback, Stiennon et al., 2020: https://arxiv.org/abs/2009.01325.
- Transformer Reinforcement Learning, Leandro von Werra, 2020, GitHub: https://github.com/lvwerra/trl.
- Combining Deep Reinforcement Learning and Search for Imperfect-Information Games, Brown et al., 2020: https://arxiv.org/abs/2007.13544.
- Distribution Augmentation for Generative Modeling, Jun et al., 2020: https://proceedings.icml.cc/static/paper_files/icml/2020/6095-Paper.pdf.
- A Wholistic View of Continual Learning with Deep Neural Networks: Forgotten Lessons and the Bridge to Active and Open World Learning, Mundt et al., 2020: https://arxiv.org/abs/2009.01797.
- Evolving Machine Learning Algorithms From Scratch, Real et al., 2020:https://arxiv.org/abs/2003.03384.
Technology Readiness Level : TRL 2
Solving UNESCO’s Educational Objectives to Support the Achievement of Education for All.
The Intelligence Foundation is looking for corporate sponsorship, foundation grants, major gifts to support Saraswati AI. The Intelligence Foundation issues receipts for charitable donations under the number 85593 8502 RR0001.
✉️ Make an impact on the world : patronage@intelligencefoundation.org
Invite link to join MONTREAL.AI on Slack to help people coordinate : https://join.slack.com/t/montrealai/shared_invite/zt-f3586fu9-TsgE5tW5b8uE3cGmtrSgMw