Introducing Project Clarify
The last decade or so of progress in deep learning enabled AI has been amazing — whether it’s strategic game play (e.g. Go, Starcraft, League of Legends), video understanding that enables self-driving cars, or AI that can accelerate protein folding simulations over 300,000 times. The foundation of this progress have been massive data sets, progressive doubling in compute power, the growth of cloud computing, the maturation of the GPU (and other accelerator) industry, and super-exponential growth in the deep learning research community.
The future of artificial intelligence holds untold promise for the enhancement of human cognitive and emotional effectiveness. The merit of this is obvious from the perspective of supporting human flourishing. It’s also clear from the perspective of any organization that employs people. For example, imagine the value of just a 1% increase in productivity to a corporation like Facebook or Alphabet. The expected value at p=0.01 (1% chance of 1%) of this in regard to a change to reported Q2 yoy increases would be $466,391 and $749,672, respectively (per quarter). Easily a low estimate and as we’ve seen with AI in other domains — the sky is the limit.
We have a plan for achieving this. One core component is having people play games that are cognitively challenging and serve as metaphors for the forms of work or other real-life performance we hope to benefit. Playing these games gives near continuous feedback on how well someone is performing. At the same time we will use proven deep learning techniques for understanding the person’s emotional and cognitive state from video of their facial expressions, measures of their cortical activity, as well as various other modalities— in short, state understanding. The combination of state understanding and game play will give us a clear basis for (1) which states to cue a person to attain and, (2) in real time, whether they are in those states. For example, helping you self-induce highly productive flow states every time you sit down to work.
We are building a platform that is vertically-integrated and modern in regard to the technologies used because we have the experience to understand the importance of this across various dimensions — the obvious ones including maintainability, extensibility, reliability, and performance; a natural choice was to build with components from Kubeflow and on Kubernetes as well as to use TensorFlow, Tensor2Tensor, and Polymer + TypeScript on the front-end. We are further building this platform to support the development and trialing of AI-first digital therapeutics — seeking to greatly accelerate the exploration and pilot study process of new designs so the field discovers many more great paradigms much more quickly and efficiently.
The Project Clarify code base is fully open source with the objective of broadly facilitating the community and can be found on GitHub with a growing list of open issues here: https://github.com/projectclarify/pcml/issues.
We’re looking forward to welcoming you into this community and helping make space for the diverse range of skills and perspectives that can form the crucible from which we will forge brilliant new capabilities for the benefit of human thriving.
Let’s build this future together.
Christopher
Founder, Project Clarify
