Yes, you did read correctly, this story is about Artificial Intelligence (AI) — and Cards Against Humanity. How they both got together, you ask? I will clarify this in a bit, but first of all let’s have a look about what AI even is…
Wikipedia defines Artificial Intelligence as
intelligence demonstrated by machines, in contrast to the natural intelligence displayed by humans […]. Colloquially, the term “artificial intelligence” is often used to describe machines (or computers) that mimic “cognitive” functions that humans associate with the human mind, such as “learning” and “problem solving”.
Some of the most common AI applications are:
- Apple’s Siri
- Amazon’s Alexa
- Microsoft’s Kinect
So, whoever is having an iPhone for example is basically an owner of their very own tiny intelligent machine. At this point I also would like to mention that — despite using a picture of a robot at the very top — robots are not necessarily AI. Robots are machines which are “programmable” and “carry out a series of actions autonomously, or semi-autonomously.” Such as in a car factory when you’re having robots to build certain parts of a car — the machine does act in a certain way and completes the required tasks automatically (or semi-autonomously), but it’s far away from having the opportunity to “learn” or “solve problems”. However, a robot can of course be programmed to have artificial intelligence.
I guess that everybody is aware of the increasingly important topic around those developments involving AI, with a lot of positive, but also many negative thoughts towards it. The potentially dangerous sides such as incidents with self-driving cars or the misuse of AI for political (and other) manipulations are discussed widely and many well-known experts warn(ed) about using AI for the wrong means:
What all of us have to do is to make sure we are using AI in a way that is for the benefit of humanity, not to the detriment of humanity.
— Tim Cook, CEO of Apple
Success in creating AI would be the biggest event in human history. Unfortunately, it might also be the last, unless we learn how to avoid the risks.
— Stephen Hawking (1942–2018)
While any potential benefits and risks should of course be taken into account at all times, I do not want to go into the details about it in this blog post, but rather talk about something fun which I came across browsing the internet and thought I should share with you all: Cards Against Humanity’s Black Friday AI Challenge!
In this epic 16-hours-battle — or “elaborate stunt to get attention and make money while overworking our employees” — , writers of Cards Against Humanity were facing AI in order to prove who is better in creating the famous cards. The “cannabis-smoking” writers had this “one chance to save their jobs from automation.” Jokes aside (for now), they actually did use an AI built on top of a neural network called ‘GPT-2’ made by OpenAI. The purpose of the network is to generate realistic-looking text through algorithms which are capable of learning. It basically attempts to predict the following words, like so:
By predicting many words, it assigns a probability distribution and uses this in order to select the next word, which randomises the whole process, making it creative and original. At first this special network was trained by OpenAI on 40 GB of internet text, which is the equivalent of 40,000 books. After this, GPT-2 had to undergo another training where it had to ‘study’ about 2,000 official cards, 25,000 internal brainstorming cards which are not included in any game, and 17,000 unofficial cards from fan-curated lists. The AI began to process — and the company began to filter out cards by looking at the format, potential banned words/jokes, similarity to other cards or just some weird stuff like “~~~~~~~~~~~~~ meh”, for example.
Here are some results of AI produced cards (I have to apply a very generous filter here, but if you’re really curious, look it up with the link I shared above):
- “Big, powerful poops.”
- “Gary’s dad, Gary.”
- “Killing, eating, drinking, and otherwise enjoying life.”
- “Giving grandma crappy things she doesn’t like.”
- “A woman who is so smart that she can predict the movements of 300 birds.”
And yes, you can actually purchase those cards :)
Luckily, the writers won in the end and sold 2% more packs than the AI, “so their jobs will be replaced by automation later instead of right now.” Of course the description of the challenge itself should not be taken too seriously — but what can be seen here is how AI can be used for some fun (marketing) ideas (and I really hope that the writers at Cards Against Humanity do see it the same way).
Before I am finishing this not too serious post, I would like to share one of the very funny Q&As on the original page:
