Microsoft has broken a technological milestone, after creating voice recognition software as good as humans. Microsoft said in a blog post that its new technology had a word error rate of just under 6%, just as good as a human transcriber and good enough for flowing conversation. Researcher Geoffrey Zweig said that the breakthrough was due to “twenty years of effort”, and the use of machine learning. However, the technology does still struggle in particularly loud environments and is also said to be likely to struggle with new accents. No date has been given as to when Microsoft will be including the product within its devices or operating systems.Â
