Typically, investigators write code or build robots from scratch whenever starting a new AI or robotics project. This leads to specialized systems brittle to novel challenges, and lack of reproducibility. Or, off-the-shelf software such as TensorFlow is employed, which restricts work to well-studied phenomena such as synaptic plasticity. These efforts traditionally have difficulty integrating tightly with efforts in biology.
To prove this cycle can be broken, we plan to develop and deploy an open-source, continually running, cloud-based code base in which increasingly protean machines (robots and computer-designed organisms) and protean algorithms (meta learners, architecture-altering methods) are automatically designed using biological change phenomena (BCPs) incorporated as software patches. We hope that such a system could help facilitate biology-to-ALife transdisciplinary work, and help to scale up our community’s collective efforts in this regard.
To test how this approach accelerates transferal of adaptive mechanisms from biology to AI and robotics, we will host a series of workshops to initially brainstorm and then construct such a code base. Anyone willing and able to contribute code to such an effort is welcome to participate.