San Francisco, CA - Branching off the success of ChatGPT-4, OpenAI is releasing new features in an attempt to expand its user base. The new target audience, according to OpenAI CEO Sam Altman, is "...terminally online players of Magic: The Gathering." The new features boast the ability to synthesize discourse straight from a user-generated prompt. It preloads the conversation into Twitter, the site currently being spun as X by Elon Musk and people still using NFT profile pictures. This means users logging in for the morning can be immediately sucked into a dark hole of human curiosity.
Here was my first sample prompt:
Generate a small-scale controversy between 2 people, which balloons into a discussion between 10 unrelated people, all of which have never met each other before in-person. Allow for 40 replies from people who miss every single ounce of context.
ChatGPT replied back with:
Person A is excited about receiving [Magic Product] as a gift, and shares a photo online. Person B quote tweets Person A with a broad, rhetorical question about the motivations behind what causes anyone to receive [Magic Product]. Person B effectively masks a feeling of sadness which looms from elsewhere in their life, but is inflamed due to the lack of validation brought on by not receiving [Magic Product], a sadness which is buried deep beneath a facade of always having the most clever and snappy reply to anything said online.
The original exchange ripples out into tweets from unrelated participants, who take sides and post what they feel is the final word, like Martin Luther nailing the 95 theses to the church door. The controversy ends after Person C begins to complain about the existence of discourse itself.
My first attempt felt believable, a testament to the power of OpenAI's platform. But still, it lacked some of the human touch to a Magic Twitter controversy. Thankfully, OpenAI has some settings to adjust the outcome of the prompt.
I moved the Audience slider from "Ordinary, good-intentioned people who would otherwise agree on a lot of things" to "A brigade of insufferable reply guys from that side of the internet." There was even a checklist of potential replies, which covered a lot of the usual suspects. I opted for "Burner account with 2 followers", "Drama-monger who seems to always weigh in on these things", and "Former grinder using a pro tour photo from 12 years ago, with the flash turned up to the max, at a Dutch angle". These options gave our controversy the right kind of heavy, festering stench. We left this exercise with a controversy which makes you look at your phone and say "Absolutely normal one, huh."
There's still some kinks for OpenAI to work out here. We need to see it adjust for blocks and screenshots from those who have been blocked. It needs to factor lost follower count and planting users who will say "This you?" to someone who thought they finally escaped a mistake they made on the internet. But so far it's effective at synthesizing a new main character for us to focus on. When closing the window, it asked me to go into the woods and watch the way water flows over rocks in a stream. Must be some kind of bug in the programming.