MIT's AI model learns language with very little training

Published:
October 31, 2018
Coverage:
Digital Transformation More...
Categories:
Artificial intelligence
11638 thumb square
by Shriram Ramanathan
Truly disruptive

NLP technology has seen drastic improvements in accuracy and performance, leading to AI-enabled voice assistant products like Siri and Alexa. To date, most NLP algorithms have required extensive training with human-annotated sentences that highlight the structure and meaning behind words. Now, MIT researchers have developed an NLP model that closely mimics the language acquisition process in children. It works by observing captioned videos and associates words with objects and actions with little intervention from humans. Clients should monitor this development closely, as it can drastically improve the efficiency and accuracy of NLP, not to mention enhance its penetration in applications that use specialized jargon (such as manufacturing).

For the original news article, click here .


Further Reading

Automated machine learning – wait, are data scientists out of work?

Analyst Insight | November 1, 2018

With the boom of AI and machine learning, there has been a growing demand for data science skills in the workforce. Universities have started ramping up course offerings in this area; for example, MIT recently announced a whopping $1 billion investment in a new college focused solely on AI. However... Not part of subscription

Edge Computing

Technology | September 27, 2018

Software and hardware technologies that allow for performing computations at the location where data is generated. Not part of subscription

Neuromorphic chips: Still not ready for prime time

Analyst Insight | October 9, 2018

Neuromorphic chip architectures have been in research and development for decades but have seen a recent resurgence in interest (as shown in Figure 1) due to their potential to more efficiently – some initial studies have shown up to a 1,000× improvement – process neural networks and other ... Not part of subscription