A recent breakthrough in Machine Learning: The Generative Adversarial Network.
Estimated reading time: 1 minute
#IanGoodfellow is a research scientist at #GoogleBrain, a company formed after the acquisition of UK’s #Deep #Mind start-up for $525 #million. He is part of a team formed in the early 2010s at #Google that has revolutionised the world of Deep Learning, a subsection of #Machine #Learning. In 2014 he developed a new type of #DeepLearning algorithm: The Generative Adversarial Network (#GAN). GANs are a type of architecture that have propelled the developing of unsupervised learning forward in the past four years.
By using GANs, there is no need for high amounts of data to train #Artificial #Intelligence. Instead, these algorithms work in “pairs”. One #algorithm has the task to generate fake samples and the other must decide whether the presented sample is real or not. The generated samples are alternated with real ones coming from a dataset. This #method lets the algorithms train without supervision.
Goodfellow’s invention granted him a citation in #MIT #Technology Review’s 35 innovators under 35. He was also part of #Elon #Musk’s #OpenAI institute for a brief period. Even though his ground-breaking invention has enormous potential, Deep Mind has suffered constant losses since its acquisition. Only in 2016 they reported a record loss of $137 million.
Do you think Ian Goodfellow’s invention will vastly modify the #Technologic #Industry? What sectors of it do you think GANs should be applied to? Is it worth to keep investing millions into GAN research?