Artistic Data Science - AI Creates

We have talked about AI painting and AI making music.

But humans have created and mastered many more art forms. Now the question is, can we also teach our machine friends to create this art? Let's explore how AI participates in other various art forms.

?? Debate

Is debate an art form or not? It's debatable.

What is not debatable though, is that IBM has accomplished a rather impressive feat when they built Project Debater.

Project Debater is a Natural Language Processing (NLP) powerhouse. NLP is the ability of a computer to analyse and process large amounts of human language. NLP is a high-profile subset of Data Science, as researchers and corporations strive to build ML/DL applications that are better at understanding and responding to human speech. If you think about it, that is even harder for people sometimes! ???♂?

And IBM was not playing in building their application.

Project Debater learned from over 10 billion sentences taken from journals and newspapers, so it could understand human topics, regardless of the angle.

When given the topic of a debate, Project Debater uses Sentiment Analysis to classification models to find arguments on the topic and pick a stance - whether the ?? Debater wants to defend or oppose the topic.

Finally using DL models and Text-to-Speech, Project Debater ties together a speech from the strongest arguments and... well, debates! Then it listens to the opponent, translating Speech-to-Text and repeating the process to build another rebuttal. Mind. Blowing. ??

Check out their website and watch Project Debater standoff against a debate champion!

?? Photography

Who are these people? Oh they are all fake. Just the imagination of another General Adversarial Network (GAN) - that ML model where one neural network keeps improving until it can trick another. Did it trick you? ??

These "photos", along with millions more, are a part of the This Person Does Not Exist project. The project was created by Philip Wang, an ex-Uber software engineer, and uses the research made public by Nvidia.

The cool (and scary) thing is that, with enough training data, the same neural network can learn to mimic other things - cats, molecules and with the latest iteration - news articles... The GAN is fed news headlines from the Internet and then it imagines the content. Talk about fake news... ??

?? Sculpture

I asked the computer to close its' eyes and dream of a new form. -- Ben Snell

This is of course very figurative speech, but regardless of the method, art can be created in infinite amount of ways. And Ben thought of a new one.

In terms of Data Science, there is not much new here - a ML model trained on thousands of images of classic sculptures, from the Discobolus to the David and then recreated a new sculpture, quaintly different and familiar at the same time. Meet Dio ??

You should check out the webpage of Ben Snell, he has some cool algorithmic artworks. As for Dio - well she/he is made from the computer that made Dio. Dio is now the only agency of both the software and hardware that created it. There's something romantic about that, isn't there?

Ben said that was his way of "limiting his control over the algorithms". Makes one think - who's really in control? ?? Are you in control when you spend hours watching Douyin/TikTok videos? Are you in control when you buy something from the recommended list of Amazon/Tmall?

ML/DL algorithms have made our life so convenient, from optimizing delivery times, to our finances, to optimizing what we see online. I wonder if we could accept a reality without them at this point?

But enough philosophising! Last but not least, a personal favorite... ??

?? Cooking

This is one just makes you feel good.

Folks at MIT's How To Generate (Almost) Anything ventured out to find different areas where AI and humans can collaborate to create better, newer results.

And one of their experiments brought the team to Tony Naser of Crush Pizza, a Boston pizza joint.

A team from MIT used a Recurrent Neural Network trained by hundreds of gourmet pizza recipes online to create a bunch of new, (weird?), AI-generated recipes. A common theme, right? Give a model lots of data -> make it replicate without plagiarism. You're right, that's the essence of Machine Learning! ??

The cool thing here is that sometimes the PizzAI (OMG I am a genius, can't believe MIT didn't name it that way!) came up with ingredients that simply did not exist, such as "snipped caramel cheese" or "wale walnut ranch dressing". So what do you do then? ??

This is where the beauty of AI comes in - the MIT team had Tony, the pizza chef, and the machine work together - AI would create the recipes, and Tony would fill the gaps with his own experience.

The results? Let's just say, do not watch the full video on an empty stomach ??

---------

This ends my recent exploration into the artistic side of Data Science... for now.

I hope you discovered something new, inspiring or learned a bit about Data Science in the process! ??

If you are curious about what else is possible with Data Science, check out the Le Wagon Data Science Program in Shanghai and other cities around the world! ??

?? Love each other, humans. We are all in this together.