We are a Barcelona AI LAB solving General Intelligence.
On top of our novel patent-pending computational model of biological cognition we have implemented a new game-changing algorithm to solve General Intelligence, upscaling from motor-perception towards figurative processing.
General Intelligence will ignite a new era of truly intelligent AIs and robots. We want to lead it with our algorithm to ensure it is for the best of humankind and its wellbeing.
We pursue to solve AI grandest challenge: solving General Intelligence by understanding how the mind works and mimicking it into truly abstract-thinking machines.
These machines will be digital cognitive systems able to solve problems in the same way biological intelligences do: learn and acquire any knowledge to gain understanding and being able to reason building smart inferences about their world in order to make intelligent decisions.
Current AI algorithms are unable to genuinely understand and reason. As AIs are not truly intelligent many problems are still unsolvable. And problems currently solved by AI require brute-force, huge amounts of training data and computing power. Current AI solutions are narrow and not reusable for other problems.
Solving General Intelligence will open a new era of Cognitive Computing where truly intelligent AIs won’t need to be programmed but educated to gain genuine understanding and reasoning.
We want to lead this new era of AI, ensuring that our algorithm is available to everyone. Consequently we are not only focused on the validation and demonstration of our General Intelligence algorithm, but also on developing the most usable platforms so anyone can seamlessly use our algorithm to build their own intelligent AI, digitize their expert knowledge to cognitively solve any problem.
Our computational model of thought mechanics
We have faced the challenge of solving General Intelligence from a deep computational perspective, and always inspired by neuron-based General Intelligence living beings.
We have understood Cognition as the information processing model and computational mechanisms underlying abstract thinking phenomena exhibited by all neuron-based species through the evolution timeline.
Without involving neural networks nor symbolic approaches, we have successfully modeled the cognitive computational workings and mechanisms of living minds with a novel and unique bio-plausible, patent-pending and still undisclosed bottom-up computational model: the Computational Fractal Cognition Model (CFCM).
This model is built on top of very essential and novel approaches but aligned with specific reputed computational theories, models and authors.
As a basic premise, we have postulated that Cognition operates under nature’s fractal computation paradigm: a fundamental cognition primitive that adaptively distributes throughout self-organized hierarchical structures that dynamically and emergently upscale the primitive computational mechanisms.
Our computational model of the mind upscales initially in a foundational Computational Theory of Cognition (CTC) that models low-cognition sensorimotor processing using spatio-transitional conditionable cognitive architectures based on perception/actuation computational mechanisms and effects. Afterwards it upscales into a Computational Theory of Mind (CTM) that also solves high-cognition qualia-based phenomena using episodic and attentional abstract processing computational cognitive mechanisms and their higher cognition and intelligence behaviors.
Once these cognitive architectures are embodied and connected to any set of domains and conditioning signals they produce machines that continuously learn, abstract and infer from their connected domains.
This learning process, that can be indistinctly unsupervised or by behavioral education or a mix of both, fills these machines with unified cognitive ontologies of knowledge that model their domains finding their relevant semantics.
The resulting educated machines can make predictions and produce intelligent behaviors according to this acquired knowledge.
Our high-cognition systems implement all necessary cognitive mechanisms to be equivalent to general intelligence machines, that provided with the appropriate education and knowledge will even express human-like intelligence capabilities.
Our master algorithm
We have implemented our computational cognition model into a unique and working AI Master Algorithm: Synthetic Cognition. We have postulated and are successfully validating and demonstrating that it is the first General Intelligence Master Algorithm. Using our algorithm any low-cognition or high-cognition architecture can be seamlessly embodied into AI machines to solve cognitively any problem, no matter the required cognitive capabilities. Machines that can be embedded with goals through conditioning policies and can unsupervisedly or by education acquire knowledge from their embodied domains, continuously enriching this knowledge while simultaneously building smart inferences and producing intelligent decisions and behaviors.
Synthetic Cognition has other unique characteristics that will be game changing and will very quickly spread it as the new dominant standard algorithm of a new AI era.
Traceable & interpretable
It is not a black-box algorithm. Its unified knowledge cognitive ontologies are traceable and its predictions interpretable
Being it truly cognitive it will require much less data and much less computational power to learn and predict properly
As it has high generalization capabilities it will allow real valuable and reusable knowledge digitalization
It is universal as it is able to work simultaneously and indistinctly with different data and signal types and without any prior and/or hardcoded knowledge
The Avatar Cognition team is following a 2 year bottom-up roadmap from cognition synthesis to general intelligence.
Using Synthetic Cognition algorithm we are designing and embodying different cognitive architectures upscaling from low-cognition into high-cognition, in order to demonstrate that our computational model of though mechanics really solves the whole General Intelligence problem.
Low-cognition roadmap stage is already solved. Synthetic Cognition is working and already solving diverse and relevant AI problems (prediction, perception, reactive autonomous agents…).
In the following months we will complete our roadmap by also upscaling Synthetic Cognition into currently unsolved high-cognition mechanisms, enabling genuine attentional abstract thinking general intelligence machines (reasoning, decision making, autonomous behavioral robots, communication and language…).
The availability of working General Intelligence machines will open a new roadmap of knowledge transfer and machine education of high-cognition embodiments, many of them imitating human-level cognitive capabilities.
More on our Technology
The concept of Master Algorithm was introduced by Pedro Domingos in his 2015 book “The Master Algorithm: How the Quest for the Ultimate Learning Machine Will Remake Our World”.
The Car Racing environment is part of the OpenAI Gym platform for reinforcement learning.
Robotics is one of the major fields where artificial intelligence is used to increase productivity via the automation of repetitive processes in numerous industries.
Find more information on our blog
Co-founder, co-CEO and Chief Scientist
Inventor of our computational model of the mind. From an out-of-the-box-perspective, he has been passionately devoted for 25 years to the computational understanding, analysis and modelling of nature’s cognitive processing model.
Co-founder and co-CEO in charge of business and operations
Passionate about cognition, AI and sci-tech innovation. He pursues to change AI for good with our algorithm, and ensure that this happens also for the best of humankind.
UPC BarcelonaTech Telecom School EE Professor. AI PhD and researcher in several topics including AI-cognitive Future Internet.
Senior lawyer with long experience in technological startups advisement, from funding towards M&A.
Josep Lluís Sanfeliu
Nearly 20 year experienced investor, board member and advisor in deeptech start-ups with special focus in bio-tech and health-tech.
For 14 years CEO of Scytl, the e-voting global reference. He is now the CEO of the successful online travel agency Exottica.