COIN: Bridging Human and Machine Learning for the Needs of Collective Intelligence Development
- 0 Collaborators
Training is the major need of the Artificial Intelligence (AI), which humans can provide at this stage of the AI evolution. In the human world, these issues are addressed by education, which is an important enabler of human development implying expansion of capabilities (also creative ones) and freedoms. Likewise, a modern AI system must be well trained. We argue that (deep) learning for a machine is a dynamic, evolutionary process, very similar to a traditional higher education, however, with new challenges and features. Recently we introduced a concept of the University-for-Everything or “deep university”, which can be launched as a creative, evolving and collaborative training environment for artificial (AI) and hybrid (Human + AI = Collective Intelligence) cognitive systems (“students”). Such university will provide a “student” with the information resources and training methods for the increase of the self-awareness and autonomy by deep-learning-driven self-development of capabilities to be curious, to ask questions (formulate queries) intelligently and creatively, find answers and make decisions. Collective intelligence (COIN) project is concentrated on performing studies needed for launching the University-for-Everything, which is the environment for collaborative development of collective (human + machine) intelligence. Generic goal of this research is bridging the gap between human learning and machine learning for their mutual benefits and coevolution. We aim to prove (theoretically and experimentally) the hypothesis that the selection, grouping and ordering the training material matters for both human and AI learners and does it in a similar way. For that we are going to study a variety of ontology-driven and deep-learning-driven metrics capable to rank (order) samples as well as groups of samples for learning towards optimal efficiency of the training process; to make synchronous tests by training deep neural networks (ensembles with different configuration and pre-training) and human groups and assess the performance of training depending on the order of the inputs; to simulate the collective intelligence groups by combining the deep neural networks with the autonomous “clones” of the particular humans according to our “patented intelligence” (PI-Mind) technology. The result of the research is a theory and a pilot of a training space, where students together with AI will be able to learn how to address complex problems collaboratively. ...learn more
Project status: Concept
Intel Technologies
Other
Overview / Usage
COIN project is concentrated on performing studies needed for launching the University-for-Everything, which is the environment for
collaborative development of collective (human + machine) intelligence.
Generic goal of this research is bridging the gap between two domains (human learning vs. machine learning) for their mutual benefits
and coevolution. The objective of this research is the pilot study with questions below:
- Which of the machine-learning techniques would be reasonable to apply also for human education;
- Which of the human-learning techniques would be reasonable to apply also for AI training or machine learning;
- What would be a suitable integrated collection of the human-learning techniques and the machine learning techniques for training the
collective intelligence (groups of humans and AI systems);
The goal is to prove (both theoretically and experimentally) the hypothesis that the order of training material matters for human and AI
in similar way.
https://m3.jyu.fi/jyumv/ohjelmat/it/en/itkm011-intel-workshop-in-collective-intelligence/211118 link to the video about this project
Methodology / Approach
As our generic goal is bridging the gap between these two domains (human learning vs. machine learning) for
their mutual benefits and coevolution, we are looking for the answers for the three generic research questions:
Which of the machine-learning techniques would be reasonable to apply also for human education; what
would be the expected benefit and impact of it; and what would be the process?
Which of the human-learning techniques would be reasonable to apply also for AI training or machine
learning; what would be the expected benefit and impact of it; and what would be the process?
What would be a suitable integrated collection of the human-learning techniques and the machine learning
techniques for training the collective intelligence (groups of humans and AI systems); what would be the
expected benefit and impact of it; and what would be the process?
The specific objective of this project is the pilot study, which is addressing the three questions above while
focusing specifically on the incremental learning and combination of the mini-batch with incremental learning.
We want to prove (both theoretically and experimentally) the hypothesis that the order of training material
matters for both human and AI learners and does it in a similar way.