Communications of the ACM
Meet the Scientist Teaching AI to Police Human Speech
Alexis Conneaus work has helped Facebook and Google build artificial intelligence systems that can understand dozens of languages with startling accuracy.
Credit: Brian Flaherty/The Washington Post
Facebook and Google have engineered artificial intelligence (AI) systems capable of understanding dozens of languages with remarkable accuracy through the efforts of scientists like Alexis Conneau.
At Facebook, Conneau and others advanced machine learning algorithms' ability to abstract language numerically, eventually training an AI model to piece through different languages concurrently; the 100-language XLM-R model was almost as accurate as its specialized single-language peers.
Conneau's final work for Facebook was on wav2vec-U, an unsupervised speech-recognition system that reads words from audio.
Conneau has helped lead research on natural language processing, and spearheaded work in AI that Facebook and others have applied to the online policing of bullying, bigotry, and hate speech.
He believes this problem can be addressed only through automation, while critics claim such innovations will just give companies more information on Web users to exploit.
From The Washington Post
View Full Article - May Require Paid Subscription
Abstracts Copyright © 2021 SmithBucklin, Washington, DC, USA
No entries found