A viral chatbot claiming to be powered by AI has a much more human secret. Every response originally came from one increasingly exhausted man.
ChatTJB presents itself as the “leading chat interface powered by AI,” but its billboard contains an important disclaimer. AI stands for “average individual,” not artificial intelligence.
That individual is Tucker Bryant, a 32-year-old artist, speaker, and former Google employee who manually responds to users through a website designed to resemble ChatGPT.
“Picture this: ChatGPT, but on the other end it’s just me giving bad advice. Manually,” Bryant explained in a video.
The project’s website describes ChatTJB as “artisanal intelligence, handcrafted by a single human being,” powered by what it calls a “single-operator large language experience.”
ChatTJB overwhelmed by 5,000 prompts an hour #
Bryant launched the chatbot alongside a $6,000 billboard in San Francisco on July 27. Despite the ad being partly obscured by trees, photos began spreading online and sent users flooding to the site.
ChatTJB initially received between 50 and 100 questions per day. That climbed to roughly 1,000 daily prompts before a viral Instagram video caused traffic to explode to as many as 5,000 requests per hour.
Bryant told Futurism he had been answering every message until the volume became impossible to manage. Users asked for everything from waffle recipes and dinner ideas to relationship advice and hand-drawn images.
Though some users just used the service to troll, Bryant said ChatTJB “is in some ways a troll” itself, so their response is reasonable.
Related #
Some users even reminded him to drink water and take breaks after learning a real person was behind the responses.
“Some people are just looking to chat, asking about what to make for dinner, etc. And some people are coming with their personal life stuff, and it’s been really sweet to engage with them,” he remarked.
Bryant created ChatTJB to challenge how readily people trust polished answers from AI platforms, even when those responses may be inaccurate. He stressed that the public art project is not anti-AI, but aims to make users think more critically about the information chatbots provide.
More than 1,000 people have since offered to help answer messages. Bryant is now carefully testing approved volunteers and moderation tools, meaning ChatTJB may not remain a single person operation forever.
The stunt follows another AI trolling experiment in which a developer hid an Old English prompt in his LinkedIn profile to make automated recruiter messages sound medieval.