5 Bizarre and Hilarious Ways Data Science is Being Used in the Real World

When most people think of data science and machine learning, they picture complex algorithms crunching serious datasets to tackle weighty business and societal problems. But it turns out, data science can also be used for much weirder and more lighthearted purposes. As AI capabilities have advanced in leaps and bounds in recent years, creative technologists and data scientists have found some truly bizarre applications that showcase the full scope of what‘s possible.

In this post, we‘ll take a tour of five of the funniest, most surprising real-world examples of data science being used in eye-opening ways. From writing weird AI-generated stories to creating trippy art and predicting reality TV outcomes, these applications may not solve world hunger – but they certainly make for fascinating conversation starters! Let‘s dive in.

1. AI Dungeon: A Never-Ending Story Written by GPT-3

You may have heard of GPT-3, the massive language model created by OpenAI with 175 billion parameters. It can write shockingly human-like text when given a prompt. Nick Walton, an indie game developer, took GPT-3 and ran with it, creating an interactive fiction game called "AI Dungeon."

In this choose-your-own-adventure style game, the player types in actions and dialogue, and the AI continues the story from there, generating paragraphs of rich narrative prose. The results are frequently hilarious, non-sensical, and full of plot twists. For example:

"You are a knight living in the kingdom of Larion. You have a steel longsword and a wooden shield. You are on a quest to defeat the evil dragon of Larion. You‘ve heard he lives up at the north of the kingdom. You set on the path to defeat the dragon and walk into a dark forest. As you enter the forest, you see"

> You slash at the trees with your sword as you walk through the forest.

*"Slashing at the trees with your sword, you accidentally chop down an entire tree which falls on top of you, crushing you instantly. Your quest has come to an abrupt end.

THE END."*

Oops! The AI took that story in an unexpected direction.

The game took off and went viral as people discovered how weirdly entertaining it could be to "co-write" a story with an artificial intelligence. AI Dungeon has been played millions of times and people have posted some of the most amusing story excerpts on social media.

What‘s fascinating is how the AI can handle the incredible diversity of prompts and plot directions that players throw at it, while (mostly) maintaining coherent story structure and language. It shows the incredible potential of mega-language models like GPT-3 for creative storytelling, even if the stories often take absurd and ridiculous turns. AI Dungeon is a great example of data science being used for pure entertainment value.

2. This Person Does Not Exist: AI-Generated Headshots

Artificial intelligence can now generate photorealistic pictures of human faces. And we‘re not talking about simple Photoshop jobs – these are images of "people" who do not actually exist in the real world at all. They were dreamed up entirely by AI.

The website ThisPersonDoesNotExist.com, created by Uber software engineer Phillip Wang, uses cutting-edge machine learning to generate a unique facial image from scratch every time you refresh the page. The AI has been trained on a huge dataset of real human faces so that it can create novel faces that look convincingly real, down to the pores and individual strands of hair.

Some of the AI-generated headshots do have tell-tale glitches, like asymmetrical glasses or teeth. But at first glance, most are indistinguishable from photos of actual people. It‘s a showcase of how powerful generative adversarial networks (GANs) have become at understanding and creating images.

The same AI technique has also been used to generate fake Airbnb room photos, fake anime characters, fake memes, and more. Aside from being a fun tech demo, this AI capability has concerning implications around disinformation, fake social media profiles, dating scams, etc. As the AI gets better, it will be increasingly difficult to tell real from fake online.

3. Botnik Predictive Writer: AI-Generated Parody Scripts

What happens when you feed a massive repository of Seinfeld scripts, Harry Potter books, or Savage Love advice columns into a predictive keyboard AI app? You get some hilarious mashup parody content that sounds kind of like the source material, but is utterly nonsensical.

That‘s the premise behind Botnik, a creative tech community that trains machine learning models on highly specific datasets (e.g. Scrubs scripts, Willy Wonka movie scripts, etc.) and then uses the results to generate weird new scripts that blend recognizable phrases and tropes from the source material into surreal new arrangements.

For example, here‘s an excerpt from the AI-generated Seinfeld script entitled "The Poot Flat":

"JERRY: I have to get my tooth fixed. It‘s gonna be a barge endeavor, but I don‘t want to end up with a deep cavity. Hey, did you hear about that guy who overdosed on placebos?

GEORGE: The most important thing is to take it sleazy.

ELAINE: I‘m sorry, I‘m just not in the mood.

KRAMER: Elaine, you know I‘ve suffered from clavicle cancer!

ELAINE: Oh god. I forgot."

Predictive text and RNN models like those used by Botnik look at patterns in the source text and try to mimic the style while coming up with novel sequences. The results are uncanny – it really does "sound" like a Seinfeld scene, with the characters‘ speech patterns and go-to phrases. But the dialog itself quickly veers off into absurd nonsense land, to hilarious effect.

Botnik‘s viral hit "Harry Potter and the Portrait of What Looked Like a Large Pile of Ash" made the rounds a few years ago, proving how entertaining it can be to let an AI run wild with well-known fictional universes. While not the most sophisticated use of data science, Botnik‘s experiments are a good example of how AI can be used for parody, satire, and absurdist humor.

4. MLBN: Using Machine Learning to Predict Baseball Games

Major League Baseball is one of the most statistics-obsessed sports leagues out there, with over a century of detailed historical data about every pitch, hit, and catch. As you might expect, this makes it a prime candidate for applying machine learning to predict the outcomes of future games and player performance.

Back in 2015, the MLB launched its own Statcast AI system that uses radar and computer vision to capture data on every play. This provides incredibly granular details like a batter‘s swing angle and the spin rate of pitched balls. Data scientists have used this rich dataset to build predictive models for things like pitch framing, base running, and defensive positioning.

But more recently, data scientists have taken MLB game prediction to new levels. The MLB Network launched a new broadcast in 2022 called "MLBN" which uses live predictive stats powered by Google Cloud. Before every pitch, the AI flashes a "probable pitch type" based on factors like the game situation, batter and pitcher stats, and stadium. It also predicts hits and outs before each play.

Sports broadcasters have long leaned on stats to fill air time and provide extra context for fans. But never before have we seen this level of real-time, AI-powered predictions during the game. It remains to be seen how accurate the MLBN model is over a full season. But it‘s a compelling use case for practical machine learning in sports that directly enhances the fan experience.

The MLB plans to make the live predictions available in its mobile app and online gameday experience as well. What‘s more, Google Cloud intends to make its sports prediction system available as a commercial product to other leagues worldwide. So we may soon see AI-powered play predictions in everything from the NFL and NBA to the Premier League and IPL. It‘s a whole new world for sports nerds to geek out over.

5. Analyzing 5 Years of Kardashian Headlines

Love them or hate them, the Kardashians are one of the most talked about families on the planet. With their long-running reality TV show, tabloid headlines, and massive social media presences, they produce a firehose of pop culture content. But what can we learn by applying data science to study the Kardashian media machine?

That‘s what data journalist Mona Chalabi sought to find out. She collected over 30,000 headlines about the Kardashians from the past five years from various gossip sites. Using natural language processing techniques, she and her team parsed the text to detect trends, sentiment, and topics over time.

The analysis revealed some surprising (or perhaps not so surprising) insights:

• Kim Kardashian is mentioned in over half of all the analyzed headlines, cementing her status as the most famous sister
• Words like "butt", "booty", and "curves" appear very frequently in headlines about Kim, reflecting society‘s obsession with her body
• Kylie Jenner saw a massive surge in headline mentions starting in 2014 as her cosmetics company took off
• Headlines about the Kardashians overall grew 50% from 2015 to 2019, showing their enduring star power
• Machine learning sentiment analysis showed that Kourtney had the most positive headlines while Khloe had the most negative

Chalabi acknowledges that studying Kardashian gossip headlines is not the most serious or impactful application of data science. But it demonstrates how the same techniques used to analyze things like stock prices, customer reviews, or political news can also be used to derive insights from our celebrity-obsessed culture.

By turning the Kardashians‘ public lives into quantifiable data points, we can track the family‘s cultural relevance over time, see how the media narrative about each member has shifted, and even pick up on subtle linguistic trends. The results paint a fascinating picture of fame and tabloid journalism in the social media era.

While it‘s far from the most important use of data science, analyzing Kardashian headlines is a quirky way to showcase the wide applicability of machine learning and NLP for deriving insights from text, no matter how seemingly superficial. And admit it – you‘re at least a little curious what Kim‘s most popular butt-related adjectives are. Data science can tell us!

Closing Thoughts

As these five examples illustrate, data science doesn‘t always have to be serious business. The same techniques and tools used to forecast stock prices or optimize factory operations can also be used for artistic expression, cheeky humor, or silly experiments.

What unites all these use cases is the core premise of data science: using computing power to find patterns in large datasets and extrapolate something useful (or at least interesting) from it. Whether that‘s using AI to write a wacky Harry Potter fan fiction or to predict where a baseball should position itself, the fundamental machine learning concepts are the same.

This creativity and dynamism is part of what makes data science such an exciting field to work in. As cutting-edge AI capabilities emerge from research labs into accessible tools and products, data scientists and ML engineers are becoming not only business strategists but also disruptive creative forces. They can train a language model on their favorite TV scripts, build a predictive model for their fantasy football league, or generate trippy AI art to post on Instagram. The possibilities are endless.

This matters because it inspires more people from diverse backgrounds to get into data science and experiment with machine learning. You don‘t have to be working on a big, serious enterprise problem for it to be a valid DS/AI project. The bar for tinkering around and building fun stuff is lower than ever. And who knows – your weird AI art generator may just evolve into the next big creative startup.

Of course, using AI/ML for humorous ends doesn‘t make it any less potent. All the examples we‘ve covered, from AI Dungeon to ThisPersonDoesNotExist, also hint at more serious implications around things like content moderation, misinformation, copyright, and more. Like any powerful technology, AI can be used for good or for ill, for high-minded or base purposes. Data scientists will have to grapple with these issues as the field moves forward.

But in the meantime, it‘s refreshing to see data science being applied to the lighter side of things and giving us a chuckle. Whether you‘re a seasoned ML researcher or a beginner building your first model, embracing the creative, artistic, and even weird possibilities of artificial intelligence is a great way to learn and stay engaged with the tech. So go ahead – plug your favorite song lyrics into GPT-3 and see what strange tune it composes. You‘re officially a real data scientist now.