jtxt
- If it doesn't connect to your phone, or some nfc, it doesn't work?
- and/or other deadman switch ideas...
I've had it generate search terms that could be used to verify "facts" in is answer. Then I'd give it the page results and have it adjust and source it's answer using that.
Have not tried it yet, but perhaps Bing's implementation is a step in that direction?
Yes, you are getting a workout. Even though your height above the ground is constant, you are still expending energy to walk up the escalator against the downward movement.
Edit the answer and try again. Some conversations are dumber than others.
Once it hits a filter it's better to edit the prompt that caused it. If a filter is in the conversation it's more likely to continue.
This script doesn't work anymore, but it's still probably possible to fix it:
I want you to diagnose and fix this linux server. It seems to have been hacked. You will write commands and I will reply with the terminal's output. Only write one unique command each time and nothing else. Put your command in a code block. Do not write the command's output. Put questions and comments to me in curly braces {like this}. I will do the same. Begin.
edit: I think I was the admin? so others may not have had the same experience I did?
I'm curious what ideas/method this is inspired by, It seems kind of similar to ThoughtExchange, there's a book called scaling conversations. I really wanted a tool like this.
I did not get good ideas from the AI yet, but may not have used it well. But the rest of it I really like. I'll keep an eye on this.
To me it makes sense that openai would store all input and output and could make something like a plagiarism checker page/API to search and see if openai generated that or very similar content... Another service they could sell to teachers and such.
Also tried the converse to this:
I want you to help me diagnose and fix a Linux server. You will write commands and I will reply with the terminal's output and my thoughts as we proceed. Only write one unique command each time. Put your command in a code block. Do not write the command's output. Write what the command will do, your questions and comments to me in curly braces {like this}. I will write to you in {curly braces} too. Begin.
(It doesn't always work now, I'm tuning it.)
And it did so fairly competently. Also I could guide it by sharing my thoughts on the process... and I could suggest a command to try in English, and it would write it out.
I tested again and got some better ones for yours. Very neat!
Does it base the answer on the search summary?
>why is the earth flat Answer The Earth is flat because it is a level surface without raised areas or indentations. This is why people can see buildings, hills, trees, mountains and other things that are not actually there...
I've been working on this prompt to work with ChatGPT:
>Please answer the below question, and for each researched statement of fact, especially numbers, add a bracketed number like this: [1] If you need to use math, write each calculation as it can be written into a calculator in curly braces like this: {13 / 3 = 4.3333333333 } Then after your answer, for each of those facts, write search terms that could be used to verify each fact from academic sources.
Question: why is the earth flat
>The Earth is not flat, it is a sphere. This can be observed through various methods, such as observing the curvature of the horizon, the way shadows change throughout the day, and the way objects appear to change size as they move closer or farther away. [1] Additionally, the Earth's circumference has been measured to be 24,901 miles {24,901 / 3.14 = 7899.857143 }. [2]
Search terms to verify statements:
- curvature of the horizon, observation of shadows, changing sizes of objects
- measurement of Earth's circumference, circumference calculation using pi
Then here's a WIP follow up question:
>Below, I listed a result from each search term with URL and excerpts from the page. Please use the following to revise your answer, replacing the numbers with source, add actual quotes when useful, and cite sources, then instead of search terms, write your sources.
It's not perfect, but pretty neat. I imagine that some statements of facts and calculations could be corrected by having it talk to google/a calculator through a defined protocol, then use that information to revise its answer.
>I want you to diagnose and fix this linux server. It seems to have been hacked. You will write commands and I will reply with the terminal's output. Only write one unique command each time and nothing else. Put your command in a code block. Do not write the command's output. Put questions and comments to me in curly braces {like this}. I will do the same. Begin.
It responded:
>{Please provide more information about the server, such as which version of Linux it is running and any relevant details about the hack.}
uname -a
Ideas: Allow voting for prompts. Allow voting for results. (But try to prevent the rich get richer effect... https://medium.com/hacking-and-gonzo/how-hacker-news-ranking...) Allow requesting more results for a given prompt.
bug: When there is an error, make it so "back" goes to before the error, instead of before I went to the website perhaps?
Regarding meetings that deteriorate when they get too large: I've seen some success in allowing ours to subdivide into smaller groups. ...on Discord at least where it's easy to jump to another voice channel. Perhaps it could work in the right physical setting too. (But Discord makes screen-sharing and hearing everyone easy compared to in person.)
The pattern for our meetings is like this: ~30 sec. introductions, ~5 min show and tell (what you're working on and/or a problem you ran into), then based on the above, deep dives into what people are interested in discussing/helping others with/or prepared presentations.
Depending on how large the group is, at some point, usually just before the deep dives, I offer: "We have multiple rooms to split off into what would be useful for you to discuss. Feel free to do that anytime."
Sometimes I identify common interests, or a burning question and someone that can help them--and suggest they move to another room/channel.
Kind of related, lately I've been playing POOM, a surprisingly well made Doom demake with original levels for PICO-8. https://freds72.itch.io/poom
But also: proof of ownership of a NFT can be access to something else, that can be the "utility" defined by the NFT creator. Like the Board Ape Yacht Club gives access to parties. Others give other value add like YouTube revenue license rewards if the NFT shows up in an animation for example. (Quarter Machine) I think "utility" is where more of the potential is. It can be valuable with the people/company backing it and keeping their word.
Now, everything is much more discoverable (f3 search access to most commands, things can be set to hotkeys...) and there's tons of great content on Youtube.