Sign In

Communications of the ACM

ACM TechNews

AI Software Learns to Make AI Software

View as: Print Mobile App Share:
Hands drawing each other.

A number of research organizations are working to create artificial intelligence systems capable of developing machine-learning software.

Credit: M.C. Escher

Several research organizations, including Google Brain and DeepMind, are working to create artificial intelligences (AI) that can in turn develop machine-learning software.

In many cases, the results coming from machines programming other machines match or exceed work done by humans.

If self-programming AI techniques become practical, they could increase the pace at which machine learning is adopted throughout the economy without requiring more machine-learning experts, who already are in short supply.

One set of experiments from DeepMind suggests self-teaching methods could alleviate the problem of AI software needing to consume massive amounts of data on a specific task. Researchers challenged their software to create machine-learning systems for collections of multiple, related problems. The software produced designs that demonstrated an ability to generalize and adopt new tasks with less training.

A team at the Massachusetts Institute of Technology (MIT) plans to open source the software behind their experiments, in which an AI designed deep-learning systems that matched systems made by humans on standard tests for object recognition.

However, these techniques require extreme computer power and are not yet viable replacements for machine-learning experts.

MIT Media Lab's Otkrist Gupta believes companies will be motivated to find ways to make automated machine learning practical.

From Technology Review
View Full Article


Abstracts Copyright © 2017 Information Inc., Bethesda, Maryland, USA


No entries found

Sign In for Full Access
» Forgot Password? » Create an ACM Web Account