Google search has made some advances over the past years – from auto-complete, to image search and natural language understanding. which lets Google figure out what you want no matter how badly you spell it or how you phrase it. At their Google AI and Machine Learning event held in South Africa on Tuesday, the company showcased some of their latest advances in Artificial Intelligence and Machine Learning.
Google announced that the Google Translate app, which is in 103 languages, is now available in Swahili, Zulu and Xhosa.
Speaking at the event, Blaise Agüera y Arcas Scientist for Machine Learning at Google, said, “Machine learning gives us an advance of understanding. There are certain things that machine learning enables us to do, like translation, We can now translate from one language to another without using rules. These new neural techniques ingest a sentence and then hallucinate the sentence in another language”.
Users can now have a conversation with search via the Google Assistant. The Google Assistant uses a variety of machine learning technologies to provide a better experience, from speech recognition to better ranking the information that comes back to answer your questions.
You can talk to the Google app on your smartphone – speech recognition turns the sounds into words and natural language processing helps understand what you mean. In the Google app on Android in the US 20% of searches are now by voice, and the Google app understands speech in 55 languages.
Search uses RankBrain, a ranking signal that uses deep learning to improve results. A very experienced ranking engineer working on search for a decade can guess which results users will favour with ~70% accuracy; RankBrain can do it with ~80% accuracy. And measured by user testing, RankBrain is the 3rd best-performing of all our major ranking systems.
Gmail and Chrome use machine learning to protect you from spam and malware. With neural networks, Gmail now blocks over 99.9% of spam, including some never-before-seen spam.