Facebook AI learns human reactions after watching hours of Skype
5 September 2017
People use their faces to communicate in subtle ways. A Facebook conversational bot trained on Skype chats had reactions indistinguishable from human ones
-
Subject
-
Section
-
Type
-
Topics
-
Author
-
Issue
-
Premium
Fatal AI mistakes could be prevented by having human teachers
30 August 2017
Artificial intelligences make many mistakes as they learn, which could be dangerous in driving or healthcare. Now human teachers could help them avoid errors
AI artist conjures up convincing fake worlds from memories
14 August 2017
An algorithm trained on real streets can create realistic-looking imaginary ones, and could one day be used to generate highly realistic video game worlds
DeepMind AI teaches itself about the world by watching videos
10 August 2017
AI usually relies on humans to label the world for them, but a new system from DeepMind learns to recognise images and sounds by matching up what it hears with what it sees
Sneaky attacks trick AIs into seeing or hearing what’s not there
27 July 2017
Hackers could fool driverless cars into ignoring stop signs by inserting hidden noise into images – and we don’t have a way to stop them yet
Donate your voice so Siri doesn’t just work for white men
26 July 2017
Voice assistants can struggle with accents outside their test base of white, male users. Mozilla wants samples to create systems that can handle diversity
AI suggests recipe for a dish just by studying a photo of it
20 July 2017
An algorithm trained on over one million online recipes can tell you what's in a dish and how to make it
AI coach will train hopeless chatbots to pass the Turing test
14 July 2017
A neural network trained to rate how human-like an AI is could turn Alexa and Siri into much more convincing conversation partners
Neural network poetry is so bad we think it’s written by humans
7 July 2017
A machine learning algorithm trained on thousands of lines of English poetry can now mimic different poetic styles - but its verse doesn’t quite ring true
Peering inside an AI's brain will help us trust its decisions
3 July 2017
A tool that reveals which aspects of a task an artificial intelligence is focusing on will help us understand why machines can be tricked
