Artificial Intelligence (AI) has over the past few years become an industry buzzword for next-generation machine learning and robotic platforms. But most current AI platforms are heavily reliant on training AI systems to recognize and identify patterns based in human data input. Fledgling AI startup Cogitai is trying to advance AI capabilities with a self-learning AI platform, called Continual™. The company is now making this AI platform available for businesses to put AI into action.
Founded in 2015, Cogitai backed by Sony and is managed by a team of executives with extensive experience in AI and machine learning, including CEO Mark Ring, considered a pioneer in reinforcement learning. In a recent PD&D interview, Ring discussed the company’s vision for self-learning AI.
Ring explains that most AI platforms currently operate under what he calls “supervised learning,” where a fixed data set is matched with labels defined by humans. The AI learning algorithms rely in finding patterns and learning to recognize them.
While supervised learning AI platforms are suited for pattern recognition tasks such as detection, they lack the more complex decision making intelligence required in tasks such as robotics and video games, Ring notes.
Enter self learning AI, which is the basis of Cogitai Continual™ platform. Here, AI algorithms “self-learn” through trial and reward. The reinforcement learning process incorporates a feedback loop involving the client. Through observation and learning, the Continual™ platform builds up its knowledge base to guide future decisions.
According to the company, the Continual™ platform allows human training of the AI to focus and speed the learning process, and automates nearly repetitive processes. The platform includes client libraries in Java, Javascript, Python, and others that communicate with Continual™, and lower-level APIs for machine learning, reinforcement learning customization as needed.
Dennis Crespo, vice president of marketing and business development for Cogitai, says that the company is now working with early adapters wishing to use the platform for their business or manufacturing processes. As an example, Crespo says an automotive manufacturer is trying to adopt the platform to implement an engine management control system based on predictive behavior rather than preset limits, to optimize engine performance and fuel economy.
Cogitai CEO Ring notes that the Continual™ platform could evolve into what he considers a third-generation AI platform based on continual learning. At this stage, the AI platform will use its learning skills and knowledge to make even more complex decisions than at present.