Home - Research - Introducing Google’s New “Thought Vector” Algorithm

Introducing Google’s New “Thought Vector” Algorithm

Introducing Google’s New “Thought Vector” Algorithm

Google, who are forever on the look-out for the next big development are working on a new algorithm. Introducing ‘Thought Vectors’ to aid them in their development with a number of different projects, most notable the Android version of Siri, which is yet to be named.

Professor Geoff Hinton, who has been an employee at Google for a couple of years, is one of the people at the forefront of the project. He has said previously that Google are “on the brink of developing algorithms with the capacity for logic, natural conversations and even flirtation.”

This sort of technology is a step ahead of what is currently available on the market.

What are Google looking to accomplish with their Thought Vector algorithm?

As mentioned, Google are essentially looking to create their own smarter version of Apple’s voice recognition software, Siri.

Google’s version, which is yet to be officially named, has been predicted by professionals in the industry to ‘have more common sense’ than what is currently available. It will do this by determining different meanings that humans use and encode these thoughts as a sequence of numbers.

Hinton is said to believe the approach will help to crack two of the biggest challenges in technological intelligence; mastering natural, conversational language and the ability to make leaps of logic.

Hinton painted a picture where in the future, people will talk to their computer both for extracting information and even for fun. He said “There is no reason why it shouldn’t be like a friend. I don’t see why you shouldn’t grow quite attached to them.”

Hinton also detailed that he can see a “plausible path” from what is currently available to a more well-informed version with something close to a human-like capacity for logic.

“The idea that thoughts can be captured and distilled down to cold sequences of digits is controversial, there’ll be a lot of people who argue against it, but there is no reason why not.”(2) Scientists have made significant progress in the past two years, this could be a massive breakthrough in the world of technology.

Recent developments show signs of life in the industry

Recent developments in the industry have been made by Richard Socher. Socher, who is an AI scientist at Stanford, developed a programme called NaSent. He taught his programme to recognise something close to human sentiment by teaching it 12,000 sentences from the Rotten Tomatoes website.

Where did the desire to improve the market come from?

The initial motivation for Google’s new development was to improve Google Translate, which often receives a lot of negative feedback.

Google Translate uses dictionaries and previously translated documents to find typical phrases that are used in translation. This provides what one might describe as a ‘rough translation’, but does not factor in natural conversation.

How will the Thought Vector algorithm work with Google Translate?

  • Initially each work the English language will be given a set of numbers.
  • A sentence will be looked at as a path between words, which can be broken down into its own respective set of numbers.
  • The “thought” will be seen as the bridge between the two languages in translation.

It sounds complicated, but essentially these numbers will be transferred into the matching foreign language version and then decode back into a new path. The key with the whole project is to work out which numbers to assign to each word.

As one can imagine, this requires a deep etymological knowledge in every language and on top of that, a colossal amount of time and man hours.

At first, the translation will more than likely produce nonsense, but as each word is refined, the algorithm eventually finds the way in which humans use the words, and it therefore improves itself as it goes along.

On top of the above, Google has also recently updated its mobile algorithm. There is said to be slight noticeable changes in the search results on Google’s mobile search.

The two month old update is said to give a boost on mobile friendly pages from Google’s search engine, as it was previously seen as a massive downfall.

You can find out more about Google’s latest mobile algorithm update in the following blog, Will Google’s Mobile Friendly Algorithm Affect My E-Commerce Store?

Image Source: Creative Commons Flickr
Share this article