Artificial intelligence can do many different things for you, and Google AI is among the best. Here are the most amazing things Google’s artificial intelligence can do, from a game of pictionary, to help with healthcare, translating from pictures, cyber security, and even creating techno songs and artificial intelligence dreams.
Google AI Can Master “Go” In Three Days
On October 17, 2017, Google researchers announced that AlphaZeroGo, their machine learning, neural-network AI, had mastered the game “Go” in three days without any human training. This massive milestone is a step towards general AI for Google, because AlphaZeroGo was told only the rules of the game, and worked out all the strategy itself.
Google AI Can Play Pictionary With You
Google’s AI Experiments provide fun ways for people to interact with artificial intelligence. Here, Google’s AI will let you play pictionary, guessing your drawings based on its neural networks. The game is called Quick, Draw! The game tells you what object to draw, for example an owl, or a floor lamp, or roller skates. Then, you have 20 seconds to draw it using your mouse cursor or finger.
As you draw, the Google AI guesses, saying things like “it’s a squiggle… it’s a donut… it’s a rabbit… Oh I know, it’s an owl!” It’s fun and feels very human, and could be a great way to impress your grandparents!
But there’s more to Quick, Draw! than meets the eye. Because it’s an artificial intelligence, not just a computer program, it’s learning – always learning. So, each drawing you submit, whether Google AI guesses right or wrong, you are making it smarter, while having fun.
Google AI Can Help Doctors Treat Patients
Healthcare is perhaps the most important place where artificial intelligence can be applied, but it is also one of the most complicated. The practice of medicine requires doctors to spend millions of hours looking through medical records, analyzing case history, and memorizing best practices to determine the best course of treatment for any individual. Google’s artificial intelligence is beginning to become an important aid to the healthcare industry.
At a series of London hospitals, Google’s DeepMind is collaborating with doctors to save hundreds of thousands of hours a year on record-keeping, and perhaps most importantly, to automatically alert doctors when something goes wrong with their patient.
To do this, Google’s DeepMind developed a unique app called Streams, owing in some part to the sensitivity associated with medical records (Google itself will not have access to the patient data). The app alerts medical teams when a patient shows unusual symptoms, for instance, early signs kidney failure, allowing doctors to respond more quickly and efficiently than before.
While still in its early stages, Google AI’s applications in healthcare may be the most globally important and transformative.
Google AI Can Translate Anything From A Picture
While Google Translate has been around forever, artificial intelligence is making translation an order of magnitude better. One notable application is Google AI’s ability to translate words from a picture.
For example, let’s say you’re vacationing in Oslo and don’t speak a word of Norwegian. You’ve heard of an awesome restaurant by the Opera House, but don’t know which subway train to take. Easy – just hold up your phone’s camera with the Google Translate app and boom – train signs rendered magically into English. Now at the restaurant and don’t understand the menu? Rather than guess and point, Google’s AI will recognize the text and translate it for you.
Google’s AI-powered Translation services are currently available in 37 languages and is approaching human level accuracy on 18 million translations per day.
Google Artificial Intelligence Can Dream
That’s right, artificial intelligence can dream! Well, sort of. Google engineers train their neural networks to behave like a human mind, which bring different “layers” of resources to bear on a problem. For example, to recognize an image of a dog, the neural network might pass the image first through a layer that understands what an image is, then a layer that has an understanding of four-legged creatures, and on until it gets down to a layer that has experience with dogs (and perhaps even Yorkshire Terriers, if it’s good!)
Google engineers have then tried to reverse the process to improve the AI’s learning – they ask the AI to create an image based on a concept, so instead of giving it a picture of a dog and asking “what is this?”, they say “dream up a picture of a dog.”
The results of these “dreams” are amazing, and sometimes startling and beautiful. For example, here is a cloudy sky according to Google AI.
There are strange creatures, objects and other patterns the AI “sees” in the sky.
Here is a dream of “random noise”:
Beautiful and a bit creepy at the same time!
Google AI Can Make A Techno Song Out Of Anything
Another one of Google’s AI Experiments, released on November 18, 2016 byGoogle Engineers Eric Rosenbaum and Yotam Mann, is Giorgio Cam. Giorgio Cam is about one of the first well-developed capabilities of artificial intelligence – image recognition. However, Giorgio Cam adds a fun twist, that requires yet another AI capability.
To use Giorgio Cam, you hold your phone or computer camera up to any object, such as a water bottle, a dog, or a banana. The Giorgio Cam first identifies the object with image recognition, and then it creates a techno song. This requires the Giorgio Cam program find a rhyming word that makes sense, for example: “That is a water bottle; drink up to go full throttle.”
The kicker? AI voice speaks the words over techno beats by none other than Giorgio the Moroder, who is widely considered to be a founding father of techno music, and computer-driven music making more broadly. Rock on, indeed!
Google AI Can Identify Software Bugs And Cyber Attacks
As technology becomes a ubiquitous presence in our lives, the ways artificial intelligence can be applied to security are paramount. The internet is more complicated than ever, and the ability to find bugs and stop attacks by human smarts and hard rules is failing.
Google uses AI to secure the Android ecosystem against malware and attacks. Neural networks are built for pattern recognition. Expose an AI to a large enough data set, and it will create it’s own understanding of what fits and what is an outlier. This same technique can be applied to security, with neural networks scanning code for abberrations that could represent an attack.
It’s early days as far as AI security goes, but this may end up being one of the most important applications of Google artificial intelligence in the coming decades.