General-purpose language goes beyond AI Deep Learning
A team of MIT researchers are helping experts advance in AI deep learning and making it easier for beginners to understand artificial intelligence. The researchers referred to a novel probabilistic-programming system named “Gen”. They presented it in a paper at the Programming Language Design and Implementation conference recently. The researchers wrote algorithms and models from multiple fields. They did not write the high-performance code manually as well as work with equations. The best part is that Gen or general purpose language lets them write sophisticated models and inference algorithms.
For example, the researchers demonstrate that a short Gen program can infer 3-D body poses in this research paper. It is indeed a difficult computer-vision inference task that has applications in augmented reality, autonomous systems as well as human-machine interactions. This program includes components that perform AI deep learning, graphics rendering and types of probability simulations behind the scenes. The end result is better accuracy and speed due to the amalgamation of these diverse techniques.
It should be easier for experts to rapidly iterate and prototype their AI deep learning systems thereby increasing productivity.
The researchers also validated Gen’s ability to simplify data analytics by using another Gen program. This program automatically creates sophisticated statistical models typically used by experts to evaluate, interpret, and predict underlying patterns in data. It lets users write a few lines of code to uncover insights into air travel, financial trends, the spread of disease, voting patterns among other trends.
The earlier systems required a lot of hand coding for accurate predictions. This is different from earlier systems. Google released an open-source library of application programming interfaces (APIs) that helps novices and specialists automatically generate machine-learning systems without doing much math known as TensorFlow. The platform is helping democratize some aspects of AI deep learning. It is limited and costly as compared to the broader promise of AI in general as it is focused on deep-learning models.
Simulation engines, Statistical and Probabilistic models are other AI deep learning techniques available. The researchers wanted to combine the best of all worlds into one. This includes automation, flexibility, and speed. Intel would do this in partnership with MIT. It is also partnering on applications for Gen in aerial robotics for disaster response and humanitarian relief.
MIT researchers have worked extremely hard and tried to simplify AI deep learning for beginners or novices. It has been easy for them to bring different AI techniques together and work on them collectively.
