Solve human problems with machine intelligence
We are pioneering an approach to Artificial Intelligence using cognitive neurosciences and human learning theory.
This will allow our AI to solve complex problems on their own, in unpredictable environments, with implicit knowledge and less brute force.
We measure our performance by the improvement in human state, and autonomous improvement of the AI.
We want AI to step into a more significant role in supporting human problem solving.
The last decade has seen tremendous development in human cognitive neurosciences, and we now have much deeper understanding of how people comprehend the world, make decisions, solve complex problems, and work towards a goal.
Although AI has surged over recent years, most current approaches are still rooted in cognitive and neuroscience theories developed decades ago. This means that the current generation of AI will only be able to operate in certain conditions, with a lot of data, a defined end-state, and known rules.
Most of human life is not like this. Conditions are usually complex and unknown, with vague goals.
Humans do not need a lot of data when we step into a new or complex situation. We operate almost instinctively, using implicit knowledge. We recognize patterns, and make rapid and powerful decisions - even if the ‘rules’ are not known, and the information is weak.
We are not “input-output machines.” We do not assume there is one perfect answer to one perfect question. We are rivers of state, moving from where we are now, to a future goal-state, by optimizing our way through a complex maze.
We want AI to play a more significant role in supporting human problem solving. Human processes are now better understood by cognitive neuroscientists, and we are applying this in developing new approaches to AI.
Solving Problems for Society
Predicting events in society, predicting patterns of crowd behavior, optimizing for a globally positive outcome in a resource-constrained environment.
We are testing our systems against problems typically only solvable by a human. We prefer to work in challenging, 'messy' human environments, where rules are unknown or changing, data is poor, and goals are uncertain.
Our aim is to perceive the environment, understand the variables more implicitly, and find solutions. We do not consider the task to be achievable using traditional approaches to deep learning, which is based on decade-old science in human problem solving. In this old approach, solving a problem or making a decision is about a "search for the best solution by iterating your way through."
Modern cognitive sciences now know that this approach is inefficient, and usually only works if you know the desired end-state, and have significant amounts of data, which is usually not the case in human experience.
Managing a life
We are tackling the task of managing food and logistics in the home.
Life is not simple to run.
Research from the the US Department of Labor shows that is takes over 25 hours a week to manage a home. This is a complex task, because it needs to take into account all the underlying values and goals of a family or person; and then triage time, effort, and resources towards a future goal state.
Food preparation and planning consumes 12 hours per week, and yet less than 4% of groceries are bought online. This is because planning and optimizing food and resources is a complex problem. A personal assistant which can only respond to specific instructions or make shopping lists is unable to take on the cognitive load.
As more and more people work, and less people stay home to do this unpaid 25 hours a week work, the effect on society has been tremendous and largely negative. For example, restaurant spending has overtaken grocery spending in the USA, and this is having a negative impact on health, chronic disease and household savings.
We are piloting a novel approach in AI for health behavioral change, with the aim of shifting wellbeing at a population level
Research has shown that a well-designed personal AI companion can have a significant impact on the positive health behavior of a human, by intervening at exactly the right time, in exactly the right way, to shift behavior positively.
Human-level companionship, and filtering our world
Humans need to feel loved and noticed.
More than 25% of humans in developed countries live on their own. In the usa, over 26% of people report having no significant emotional support.
An AI system able to think, reason and talk like a human, and work towards the positive wellbeing and growth of that human, will have an important role in the future.
This will be even more the case as we move into a world of augmented reality and biosensors, where technology will have an intimate awareness of the human state, and ability to throw a skin of experience over our sensory world.
AI will need to become an ambient. They will need to be a predictive matrix, filtering the world for us, and tailoring our human experience for personal growth.
CEO/founder of three successful companies, 2 of those in AI
- Cognea Artificial Intelligence - CEO. Acquired by IBM Watson
- Mooter Search - CEO. Applied AI, venture backed, went public
- Toptots Cognitive - CEO. Still growing
Entrepreneur with 2 exits, and foundational contribution at 2 unicorns.
- LegalZoom - unicorn legal document company
- Swing by Swing Golf - millions of users
JR Smith is one of the foremost experts on the future of Information Technology, business leadership, disruptive innovation and consumer privacy and security, rights and dynamics, both within the U.S. and globally. He is the former CEO of AVG Technologies— in February 2012, he took AVG public on the New York Stock Exchange.
Michael has a doctorate in artificial intelligence from UNSW. He is a technologist, strategist, and cares about frontier innovation that can improve our lives, advance society, and address climate change.