An ex-Googler created an AI chat where AI stands for Average Individual
How it works: You ask a question, and a real person reads it, thinks about it, and writes back when awake and motivated. The service is intentionally slow (you may not get a reply for a while - days, not minutes), occasionally gets things wrong, and instead of generating AI images, it sends hand-drawn sketches. You can try it here: https://chattjb.org
Currently, ChatTJB is becoming a community project as the project got way more attention than anticipated. Anyone is welcome to join
Honestly, I love the idea! Nowadays some people blindly trust AI systems so this is a great reminder that not everything labeled “AI” is as magical as it seems. I still use google and despise people who ask AI simple things like “what should I wear today” or “provide ingredients of this product” - I harshly criticise my friends by asking them to use their own critical thinking skills. It’s okay to use AI for brainstorming ideas or organising things but asking it about the weather when you have a weather app??? Crazy
Anyway, what AI would you choose: Artificial Intelligence or Average Individual?
I heard about ChatTJB! I think it’s excellent humor, even if it’s obviously not really a practical tool.
Even with the best AI models today, nothing beats asking a human who’s an expert in the field your question is in. I think it will stay like that for a long time.
I think software should be mostly specialized and answer things only software can answer well, like provide information derived from finite data where things like opinion and perspective are irrelevant. AI today is the exact opposite of this: a blank text box where you can write anything imaginable. No wonder so many answers are either too generalized or simply incorrect!
Humans should answer human questions and software should process and handle raw information.
This is one of those few good uses of the internet.
I have really enjoyed LLMs for those questions I cannot formulate into a proper search-friendly phrase. Usually, because I do not know the terminology or jargon. Some examples from my recent history:
What are a group of bad actors often referred to as?
Help me remember the English word to describe an example that is perfect and to be used as a rubric.
Help me think of a word for the following situation/meaning:
If I assume someone has not done due diligence, or does not have context on what they are talking about, and want to address that in the following way:
“Without coming across as _______, why do you think that about the community?”
Or, when I want to know how to do something convoluted in Davinci Resolve, Google Sheets, or find a specific setting in VSCode.
Tongue in cheek: I hope the irony of this forum is not missed.
ChatTJB was created by Tucker Bryant. Apparently its gone off-line, he’s no longer sending outgoing messages, but the site is still active.
With so many AI agents out there, having a real actor reply to prompts sounds like a good idea. Until reality hits that thousands of people can message on any given day, and a single person then becomes overwhelmed.
I think update the terms of service to state that the message or a version of their message is saved to a database, with the reply from the prompt engineer, if any, associated with that entry. Identifying data is cleansed before reaching the database.
When someone asks a question or makes a request any similar past prompts become easily viewable.
A dyad relationship for exchange of information which can take days, weeks, or months potentially may not last.
I sometimes use it the other way around: asking to explain unfamiliar English terms that are not googleable. I provide the whole paragraph with that word for it to have some context. English isn’t my native language so it’s very practical for learning!
Tucker added a Pro version where he might reply but honestly I understand him and wouldn’t blame him if he didn’t reply at all. Answering so many prompts is tiring. A person must be really sociable to answer 100+ messages a day – especially if they’re repetitive.
This being said, I like your idea of using a database
