Vicarious concentrates on developing an integrated algorithmic architecture that would result in human-level intelligence in areas related to language, vision and motor control. Today, the company concentrates on developing visual object recognition and segmentation with the aim of creating common products that work effectively in various sensory areas and perform various roles.
Vicarious has adopted inductive strategy borrowed from neuroscience approaches. This model has reduced the order of magnitudes and offered less data training relative to traditional machine learning approaches. The company relies on modern software engineering technologies and aims to promote favorable work culture and codebase.
Vicarious has adopted a different approach to research, which differs significantly from mainstream deep learning and computer vision research. Generally, mainstream research has focused on various standard benchmarks, which allow researchers to enhance their methodologies and conduct comparative studies on results.
On the other hand, Vicarious has developed a range of questions and themes to guide its research. In addition, it has introduced data sets that specifically focus on those questions. However, when the company tests on standard benchmarks, it strives to cautiously note and classify any potential errors and their sources rather than aim to challenge the existing model of the algorithm.
The major shortcoming of the company’s approach is that it might not challenge many established standards within a short period. However, the company believes that effective implementation of its approach will enhance understanding and improve the performance of the system on a long-term basis.
Some of the themes that Vicarious has identified include generalization based on a limited number of training examples; unsupervised learning; generative models; shape model in vision; belief-propagation based inference; and the neocortex as a source of inductive biases and constraints.
The company’s research team leverages on achievements in machine learning and computer vision networks to develop its artificial intelligence systems.
Current Applications of Artificial Technology from Vicarious
Currently, Vicarious has used its technology to crack CAPTCHA used in differentiating between humans and computers in Web browsing. The technology relies on the human brain to solve text-oriented CAPTCHA tests. Users must type CAPTCHA on Web sites to prove they are indeed humans. Vicarious has claimed that its technology can solve CAPTCHA tests and perceive images.
Vicarious adopted Recursive Cortical Network software to crack CAPTCHA tests. This acted as an improvement in thinking machines and it would not serve nefarious roles. Therefore, if the advanced technology can solve a complex problem in machine vision by replicating human vision, then it is an important development in artificial intelligence. Vicarious has asserted this claim by emphasizing that the algorithm acts in a similar manner just like the human brain.
It is imperative to note that the breakthrough in cracking CAPTCHA could transform computer science and machine learning. The major challenge in artificial intelligence has been developing a system that can see and perceive images in the same manner as humans do. Therefore, the ability to crack CAPTCHA is a tremendous milestone in enhancing artificial intelligence.
Finally, Vicarious’ research can open new applications of artificial intelligence in other areas like medical image analysis, robotics, online Web browsing, and higher education, among others. Vicarious researchers claim that their algorithms have attained levels of robustness relative to actual human brains.
On this note, the company believes that it would integrate neuroscience and machine learning to develop a visual perception system based on the neocortex. This could imply that the technology may replicate human capabilities, such as hearing, seeing, moving, and speaking, among other functions.