Google – Gain in translation

Reply to this post

Google is not resting on its laurels when it comes to AI. 

• A vastly improved Google Translate is yet another sign of how far ahead Google is when it comes to tackling the key AI problems that need to be solved to take Digital Life services to the next level.
• Google has updated the translation engine for 8 of its most used languages (covering 35% of all translation queries) to use a new translation technique called Neural Machine Translation (NMT).
• Almost all machine translators break up a sentence into separate words or phrases and translate them independently which results in usable translation but often crucial nuances are missing.
• NMT takes both the entire sentence into consideration and the broader context which allows a more relevant translation which is then rearranged and adjusted to be more like the way a human would write or speak.
• Like all neural based systems, NMT learns as it goes meaning that it requires a lot of usage in order to improve.
• Google Translate is already one of the best machine translation systems available and hence I do not think that it will suffer from a lack of data from which to learn.
• On the articles I have tested, there has been a very noticeable improvement in the fluency of the translation compared to the previous version, although it is still not difficult to tell that it has been translated by a machine.
• I think this evolution is important because it has implications for Google’s ecosystem that go way beyond translation.
• AI is becoming increasingly important in terms of determining the functionality and usefulness of Digital Life services but there is another requirement that sits on top.
• This is the requirement to be able to understand and interact naturally with the user as well as have better insight into the data that the user generates.
• If the system does not understand the user or his data properly then even the best AI will almost certainly fail to provide a decent service.
• This is why this update is important as it speaks volumes about the edge that Google has in understanding what it is the user is asking for.
• Being able to translate naturally requires an understanding of natural speech which I think is the first requirement for a decent personal digital assistant.
• Personal digital assistants are the first real delivery of an AI driven service to the user and all of the major ecosystems are actively developing an offering.
• Hence, I am increasingly confident that Google remains far ahead of its peers when it comes to AI meaning that its services could become more relevant, useful and easy to use when compared to those of its competitors.
• This in turn, should drive more traffic to its servers resulting in greater opportunities for monetisation.
• That being said, I still think that much of this has already been discounted in the Alphabet share price and I am not keen to get in at these levels.
• I would prefer Microsoft or Baidu at these levels.