GoKawiilSign up to see the future, today Can’t-miss innovations from the bleeding edge of science and tech Email address Sign Up Thank you!
On July 27, a billboard went up in San Francisco advertising an AI service called ChatTJB. But the “AI” powering the service, as noted in a small disclaimer on the billboard, isn’t artificial intelligence. It’s a self-avowed “average individual” named Tucker Bryant, who manually answers user queries by way of a chatbot-like interface.
An “About” page for the service defines the process as “artisanal intelligence, handcrafted by a single human being.” Or, in other words: just talking to a guy.
Bryant, a Stanford graduate and former Google employee who now works as an artist and speaker, is clear that the project isn’t designed to be anti-AI. But in an interview with Futurism, he explained that he was inspired to create ChatTJB after his romantic partner introduced him to the phrase “cognitive surrender” — a term introduced in a University of Pennsylvania study which found that AI users are overwhelmingly likely to blindly trust the advice of chatbots, even when AI-generated responses are incorrect. It resonated: Bryant felt he’d personally experienced the phenomenon during his own interactions with chatbots, so he cooked up ChatTJB as a satirical way to infuse inconvenience and unpredictability back into a chatbot-like experience.
It’s “this idea that we often, or increasingly, trust LLM outputs so easily that we stop bringing our own creativity or critical thinking into the responses we receive, and over time that weakens our ability to be fully-functioning, sovereign human beings in the world,” Bryant mused. “I thought it would be fun or interesting to create an experience that looks like the LLMs we trust a lot — that I, at least, have fallen prey to the trap of over-trusting at times — but where the experience, rather than being this super futuristic thing I think we all felt the first time we saw those cool tokens streaming on ChatGPT, would be just a really average or bad experience.”
Indeed, chatbots like OpenAI’s ChatGPT and Anthropic’s Claude are simultaneously authoritative-sounding and eager-to-please, a combination that can lead to terrible-yet-seductive advice that many users accept with little to no critical pushback. In contrast, ChatTJB’s responses are unwieldy and unpredictable. Bryant’s responses might be good; they may also be terrible, or equivocal, or otherwise bizarre and helpful. (In a social media video, Bryant says that he wants ChatTJB users “to have such an inconvenient experience with something that aesthetically looks like artificial intelligence that the next time an LLM gives you what sounds like a reasonable answer, you remember the time ChatTJB responded to your tax question with a haiku about crows.”)
Like a chatbot, he can also generate images upon request— in the form of graphics that he hastily sketches, of course.
Insane how far AI has come.
At first, Bryant said, ChatTJB attracted a “slow trickle” of incoming queries. Which wasn’t all that surprising, given that the billboard itself isn’t exactly in a prime spot.
“It’s not a very good location — it’s obscured by trees, there’s basically no oncoming traffic that can see it, and the only way you’d get a clear look at the full billboard is if you’re parking a car in the lot right underneath it,” Bryant said. “But I was fine with that, and I also didn’t want to pay much more.” The billboard’s cost, he conceded, was “less than most used cars under 150,000 miles, but still a very irresponsible amount… $6,000 for the month, not including fees.”
... continue reading