Mathematica and Wolfram|Alpha creator Stephen Wolfram has launched a new project, which he claims is capable of identifying an image of almost anything - or at least having a very good try.
Integrated into his Wolfram Language as a callable function, ImageIdentify, the new artifically-intelligent platform is designed to provide data about any image it is presented. '[It] lets you ask, “What is this a picture of?” - and get an answer,
' Wolfram wrote in his announcement
late last night. 'It won’t always get it right, but most of the time I think it does remarkably well. And to me what’s particularly fascinating is that when it does get something wrong, the mistakes it makes mostly seem remarkably human.
For those who don't fancy writing something in the Wolfram Language to make use of the function, a public-facing implementation is available as the Wolfram Language Image Identification Project
, which includes a stock of carefully-selected sample images as well as the ability to upload any file of your choosing. Based on deep-learning neural network technology, ImageIdentify works well with clear and simple images which have a single, unambiguous subject, and less accurately the more indistinct or cluttered an image becomes. Sometimes its detection is razor-sharp, as with an image of Alan Turing it labelled as a notable person with a link to his biography, other times less so, as with a picture of a single-board computer which was labelled simply yet accurately as a 'device.'
'It’s a nice practical example of artificial intelligence,
' claimed Wolfram of his project. 'But to me what’s more important is that we’ve reached the point where we can integrate this kind of “AI operation” right into the Wolfram Language - to use as a new, powerful building block for knowledge-based programming.