search
top

Gen – A New AI Programming Language by MIT

ShareShare on FacebookShare on Google+Email this to someoneTweet about this on TwitterShare on LinkedIn

In the view of advancement in the field of computer science, especially Artificial Intelligence, Massachusetts Institute of Technology researchers have proposed a one of its kind probabilistic programming language – Gen which allows users to write algorithms without having to alter equations or superior code.

 

Introduction

There are a number of online courses which assure to infuse latest programming techniques relevant to the field of AI. But from the view point of a learner or researcher, it is difficult to cope up with the programming constructs and algorithms as far as AI is concerned. To come up with a solution to this problem, the paper “Gen: A General Probabilistic Programming System with Programmable Inference” was presented at the Programming Language Style and Implementation Conference. This new programming language, Gen, is designed to solve problems in the area of computer vision, robotics combined with statistics. It’s source code is publicly available to be used by variety of future developers.
 

About Gen

Artificial Intelligence brings together a variety of domains like statistics, computer science, image processing and psychology under one umbrella. Researchers from variety of fields not even related to computer science are interested in exploring and working on AI algorithms. But they find it difficult to build their own model on the top of existing algorithms due to unavailability of flexible platforms.
 

For a vast field like AI, the programming languages and the platforms to build AI models needs to be flexible as well as agile. Some of the challenges in building such platforms are:

  • Obtaining expected performance level for heterogeneous probabilistic models doing recursion in black box simulators.
  • Providing users with simplified implementation of inference algorithms with no or minimum restrictions.

 

Gen permits a problem-solver to use probabilistic programming, and therefore have a additional scrupulous approach to the matter however not be restricted by the alternatives created by the designers of the probabilistic programming system.
 

According to the MIT researchers, ancient probabilistic programming has some restrictions which causes lack of flexibility and potency. Gen addresses this drawback by permitting users to create specific models and programming constructs without requiring the manual effort to write code. This would allow researchers of all talent levels to make their own models and based on algorithms for AI.
 

Why Gen?

According to Marco Cusumano-Towner, author and PhD candidate in MIT’s department of electrical engineering and computer science, the motivation for this work came from the need of individuals with less experience in computer science or mathematics for AI. They wanted automated AI systems to be accessible to folks with less experience in the area of Artificial Intelligence. The intention behind Gen is to make the task of everyone easier, faster and more productive in creating AI systems, irrespective of their expertise in AI.
 

Features of Gen

Some of the interesting features offered by Gen are:

  • Gen constitutes of component parts that perform graphics rendering, deep learning, and probabilistic simulation behind the scenes improving task speed and accuracy upto orders of magnitude as compared to existing well-accepted systems.
  • Gen extend a high-level and extremely efficient infrastructure for inference tasks. It allows users to blend custom algorithms for numerical optimization and systems written in Google’s TensorFlow machine learning framework with generative models written in another AI programming language, Julia.
  • Gen allows users to use built-in algorithms with minimal lines of code to develop custom algorithms. This is possible due to the scalability of existing algorithms.

 

Some Pickups for Gen

Gen has already been picked up for further research and development:

  • Companies like Intel are already working with Gen for their AI research. Intel is using the language in its robotics and augmented reality systems. Intel is collaborating with MIT university to use it for 3D pose estimation from the depth-sense cameras.
  • MIT’s Lincoln Laboratory is developing applications in aerial Artificial Intelligence for humanitarian relief and disaster response.
  • Gen is also being used in the MIT-IBM Watson AI project and the U.S. Department of Defense’s DARPA’s Machine Common Sense project.

 

These picks are supported by Gen’s widespread capabilities.

Features of Gen

The systems build on Gen can perform various operations on data including predictions and readjusting probabilities.

2 Responses to “Gen – A New AI Programming Language by MIT”

  1. Jigna says:

    This is something new. Great!!

  2. C. J. Purohit says:

    I like all your posts. Specially those who are related to Artificial Intelligence

Leave a Reply

top