Simple AI Experiment
Pix2pix is a personal favorite among iMerit staff due to its pristine output. All users have to do is draw something (typically primitively) and the algorithm will guess what it is the user was trying to draw, and then generate a photo-realistic image of that same thing.
This AI experiment was
developed using a cGAN, or
conditional generative adversarial network. cGan empowers the
algorithm to comprehend the lines of an input to then generate
an accurate output.
For an interactive demo, click here.
Infinite Drum Machine, another AI experiment from Google Creative Labs, is an artificial beatmaker that uses everyday sounds to create fat beats. The model uses a technique called t-SNE that allows it to translate large audio datasets into smaller groups of similar sounds without the need for tags or descriptions.
When using Infinite Drum Machine, users can play with the sliding tags on the sound card to explore and use different sounds that help with phat beatsmithing.
NVIDIA's FastPhotoStyle comes straight from the mother of all GPU makers and is different from other AI experiments in that it can take any input image and generate a cool and stylized photorealistic output image. The model focuses on stylizing and smoothing input images and then generates the style of the reference image with results that can convince the human eye of its authenticity.
This AI experiment was developed and trained using almost a billion lines of dialogues so it can adequately respond to any question that’s input to it. It scans over 100,000 books in response to any question and attempts to find an answer from them.
Simply type in any question or statement you wish to test the model with, and the model will always come back with something interesting. It’s just like talking to a book.
Imagine playing piano with a virtual virtuoso.AI Duet gives you a virtual partner to play piano with that responds to any notes you input using your computer mouse/keyboard, or even an actual piano-style keyboard (assuming it plugs into your computer).
Like Infinite Drum machine, AI Duet is another shining example of how AI can function in a creative capacity.
This set of online word association games was created by Google using machine-learning and NLP technology. It’s a fun experiment that challenges the user to enter in a series of clue. The model then evaluates the words in play and comes back with the one it thinks is the most related. Eventually, the model develops connections between words based on the billions of conversations that are taking place across the internet.
Imagine creating terrible drawings that everyone could interpret. This free online image game is basically Pictionary with AI, in that you can basically draw whatever you like – be it stick figures, houses, suns, snowflakes – and the AI will attempt to guess what it is.
The more you play with it, the better the model gets at guessing what it is you were trying to draw.
This fun AI application essentially generates insanely photo-realistic images that you would be crazy to think were fake. But, as the name suggests, none of these people actually exist!
The faces are based on a dataset of training photos that were input by Uber engineer Philip Wang atop NVIDIA’s GAN network. It functions by pitting two neural networks against each other, with the first generating a face using the dataset, and the second neural network if the face looks realistic enough.
The results are perfect fake faces, every time.
Similar to pix2pix, AutoDraw takes any old chicken scratch drawing and turns it into a flawless, polished-line drawing. It was developed by a team at Google working in conjunction with a team of highly talented artists. The machine learning that powers AutoDraw compares the initial drawing input with a database of images and quickly comes back with a polished version of the drawing.
The more you play with it, the better the output! That’s the power of AI and ML at work.
Deepart.io can take any photo and turn it into an artistic masterpiece. The tool works by isolating highly favorable art characteristics across the web and applying them to any photo you upload. As a result, users can upload simple photos of their hometown and watch it transform into something that looks as though Vincent van Gogh himself painted it.
This tool is another shining example of how AI is putting the power of creativity into the hands of everyone.
Thing Translator allows users to use their phone camera to translate the environment around them. Say you’re traveling in France and need a sign translated. Using Thing Translator, you can point your camera at the sign and the phone will automatically translate it. This tool is powered by Google’s machine learning APIs and uses Cloud vision API alongside Translate API to help users navigate and engage within uncharted territory.
Fantastic for anyone looking to generate some slick rhymes, DeepBeat uses a massive lyric database across multiple languages to generate lyrics that are record ready. All users must input is an opening line, and the AI generates the rest of the song for them.
Scribbling Speech empowers you to draw with your words. This online AI tool was developed by Google to turn everyday speech into animated drawings. All you have to do is tell it what you want to see, and it’ll compare your words with its database of linguistics and drawings to come back with an epic drawing.
AttnGan is an algorithm that visualizes text input. It was developed by Microsoft’s Deep Learning Technology Center, and while the project has a noble goal with its text-visualization algorithm, the results are often just plain bizarre. For example, the output of a specific dataset (like dog images) created a convincing photo of a dog. But when it was fed a larger dataset with diverse images, the output looked very confused.
Still, it’s a fun little game to try your luck with.