How Neural Networks Are Surpassing Human Intelligence

In the past few years, artificial neural networks allowed researchers make a rather big leap towards the accuracy of software that can interpret images and other content, and we’re now seeing the boom in the AI industry. Yes, we are still nowhere close to the the plot of “The Matrix”. Unlike the human brain that can do and process multiple things simultaneously, nowadays robots need to be trained in a linear way, and scientists focus on teaching them to do one thing extremely well. However, deep neural networks have learned to drive cars, beat AlphaGo champions, paint pictures, evaluate websites, and help scientific discoveries happen.

Image and object recognition

Сapsule networks, invented by one of its fathers, Geoffrey Hinton, has almost halved the best previous error rate on a test that challenges software to recognize toys from different angles.

With capsules, small bundles of virtual neurons, the AI system is able to analyze an object and its position in a three-dimensional space. Using an increased amount of these capsules over various scans allows the system to better identify an object, even if the view is different than ones analyzed before. Take a state-of-the-art network and train it on ImageNet, the biggest database of labelled images and it will be able to classify objects better than a Ph.D. student Andrej Karpathy who was trained on the same task for over 100 hours.

Video games

Google’s DeepMind uses a Deep Learning technique that is referred to as Deep Reinforcement Learning. This method was used to teach a computer to play the Atari game Breakout. The computer wasn’t taught or programmed in any specific way to play the game. Instead, it was given control over the keyboard while watching the score, and its goal was to maximize the score. After two hours of playing, the computer was an expert in the game. After four hours of playing, the computer began understanding the true strategy and realized that digging a tunnel through the wall is the most effective technique to win in the game.

The Deep Learning community is in a race to train computers to beat people at almost any game you can think of, including: Space Invaders, Doom, Pong, and World of Warcraft. In the majority of these games Deep Learning networks are already outperforming experienced players. The computers were not programmed to play the games, instead they just played the games for hours and learned the rules by themselves.

Voice generation and recognition

Last year, Google released WaveNet and Baidu released Deep Speech, both being Deep Learning networks that are able to generate a voice automatically. The systems learn to mimic human voices by themselves and improve over time. When letting an audience try to differentiate them from a real human speaking, it is much harder to do than one might imagine. While we are not there yet in terms of automatic voice generation, Deep Learning is taking us another step closer to giving computers the ability to truly speak like humans do.

Another example: a deep network created by Oxford and Google DeepMind scientists, LipNet, reached a 93% success score in reading people’s lips where an average lip reader only succeeds 52% of the time.

Art and style imitation

A neural network can study the patterns in the strokes, colors, and shading of a particular piece of art. From there, it can transfer the style from the original artwork into a new image based on the analysis. There are plenty of new creative ways of applying this technique on the Internet.

DeepArt.io creates apps that use Deep Learning to learn hundreds of different styles which you can apply to your photos. While it might look like a filter, what is actually happening is the neural networks are analyzing the style of an example work, something like Van Gogh, and applying it to the photo uploaded by the user. The AI then reconstructs the photo from scratch.

Predictions

A group of researchers took 50 million Google Street View images and explored what a Deep Learning network can do with them. The computer learned to localize and recognize cars. It detected over 22 million cars including their make, model, body type, and year. This system was able to predict the demographics of each area by the car makeup. It also was able to make interesting ties based on the vehicle frequency, one example includes voter lines. Based on the analysis, if the number of sedans encountered during a 15-minute drive through a city is higher than the number of pickup trucks, the city is likely to vote for a Democrat during the next Presidential election (88% chance); otherwise, it is likely to vote for a Republican (82%).”

Website design

Today, website builders are either using deep networks or are going to use them in their work. At uKit, we have implemented AI to upgrade old websites built with from different platforms to the modern ones in no time.

One of our underlying technologies emulates an opinion of an average user about site appearance to provide the AI with the information whether the site design is good or bad.

So, companies are working diligently to improve the intelligence of neural networks while giving the end user services that can utilize AI and deep learning — and new ideas and improvements keep appearing. Check out one of them here.

--

--

--

Boosting website conversion with Artificial Intelligence! Designing dynamic landing pages based on crowd data: https://ico.ukit.com

Love podcasts or audiobooks? Learn on the go with our new app.

Recommended from Medium

Tech·nol·o·gy

ZI Will Kill Us Before AI Gets Out of Diapers.

The Enterprise AI Promise: first the machine, then the person

M.I.T. MCUNet brings deep learning to Internet of Things (IoT) by using $8 CPU 0.5

How Computers Understand Images and Videos with AI

Making Deep Fakes Mainstream

Soul Machines and Madera Residential’s technology subsidiary Quext, launch “Mia,” a fully…

Octhum: a tool to unify different Artificial Intelligence approaches

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
uKit ICO

uKit ICO

Boosting website conversion with Artificial Intelligence! Designing dynamic landing pages based on crowd data: https://ico.ukit.com

More from Medium

Introduction to AI | A Quick Overview of Artificial Intelligence, Machine Learning and Deep…

The Idea Behind Transfer Learning: Stand on the Shoulders of Giants

Debugging deep learning models

Machine Learning Requires Multiple Steps