Thought Vectors: Bringing Common Sense to Artificial Intelligence

The pace at which technology has been advancing over the last decade, is awe inspiring. While the myriad innovations have left us all overwhelmed by the ease they offer to our lifestyle, it feels equally bizarre to observe how easily we are taking a divorce from the organic while adopting the mechanical ways of living. In fact, we are letting go of the exclusivity of our most precious and unique asset as human beings- ‘intelligence’, by sharing it with computers and robots.

One of the finest examples delineating how fast machines are becoming an absolute substitute to us, is Thought Vectors, a highly advanced research conducted by Google under Professor Geoffrey Hinton on Artificial Intelligence.

What are Thought Vectors?

Thought Vectors is an attempt to develop algorithms that would enable computers to have “common sense”. Aaron Krumins in an article titled ‘Thought vectors’ could revolutionize artificial intelligence’ explains the concept with a very easily understandable example. He writes, the crux of the concept is that “by ascribing every word a set of numbers (or vector), a computer can be trained to understand the actual meaning of these words.”

Adding further, he jots, “Now, you might ask, can’t computers already do that — when I ask Google the question, “Who was the first president of the United States?”, it spits back a short bit of text containing the correct answer. Doesn’t it understand what I am saying? The answer is no. The current state of the art has taught computers to understand human language much the way a trained dog understands it when squatting down in response to the command “sit.” The dog doesn’t understand the actual meaning of the words, and has only been conditioned to give a response to a certain stimulus. If you were to ask the dog, “sit is to chair as blank is to bed,” it would have no idea what you’re getting at.”

Why Thought Vectors?

Thought vectors provide a means to change the abstract way computers understand human language. While ‘Natural Language Processing’ is being commonly use by tech firms globally, Thought vectors take a step ahead actually teaching the computer to understand language much the way we do.

The difference between thought vectors and the previous methods used in AI is in some ways merely one of a degree. While a dog maps the word sit to a single behavior, using thought vectors, that word could be mapped to thousands of sentences containing “sit” in them. The result would be the computer arriving at a meaning for the word more closely resembling our own.” As if the thought of this possibility were not shocking enough for us, Professor Hinton predicts that within a decade we could be counting computers among our close friends.

Hinton also maintains that Google is on the verge of developing algorithms with the capacity for logic, natural conversation and even flirtation. The idea of capturing thoughts and converting them to lifeless sequences of digits is extraordinary and hence, incredible. He is of the opinion that “there’ll be a lot of people who argue against it, who say you can’t capture a thought like that, but there’s no reason why not. I think you can capture a thought by a vector.” He believes that the “thought vector” will help overcome two major handicaps in artificial intelligence: mastering natural, conversational language, and the ability to make leaps of logic.

You might have watched the Hollywood movie Her that revolves around the concept of a surreal AI operating system. If you haven’t yet, you must, soon. The concept of human-like machine with real thoughts and emotions, has been a major part of popular fiction in the modern times, and with research and development being done on these lines, it might become a reality sooner than we expect.

In tandem with Thought Vectors, Google is working on ‘Inceptionism’

What is Inceptionism?

Inceptionism is an attempt to go deeper into Artificial Neural Networks(ANN), a statistical learning model which is inspired by biological neural networks- the central nervous system of an animal. The simplest definition of an artificial neural network, is provided by the inventor of one of the first neurocomputers, Dr. Robert Hecht-Nielsen. He defines a neural network as:

“…a computing system made up of a number of simple, highly interconnected processing elements, which process information by their dynamic state response to external inputs.

Google believes, although the developments in the ANN have pushed image classification and speech recognition to a relatively advanced level, the technology still nascent. Talking of the research being carried out on the subject, the company’s official blog states-

“We train an artificial neural network by showing it millions of training examples and gradually adjusting the network parameters until it gives the classifications we want. The network typically consists of 10-30 stacked layers of artificial neurons. Each image is fed into the input layer, which then talks to the next layer, until eventually the “output” layer is reached. The network’s “answer” comes from this final output layer.”

AI Dreams

In an experiment, engineers at Google turned the networks upside down and fed them with random images algorithms. The results came out pretty trippy-

Image Source


Moreover, the best part is that Google has let its AI code run for free on the Internet so that users can have their own share of “Inceptionism” fun by generating their own dreamy neural net image.

To wind up the article on a lighter note, here’s a video showcasing how celebrities look in deep dream. Check it out-

Is there any technology you would like for us to write about? Share with us in the comments below.

Have ideas to share? Submit a post on iamwire

One comment

  1. 1

    you know what. i now think this is a viable way towards AI, even strong AI… well done! i never new that mental falculty had other purposes….

Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>