Heard about prompt poet by character.ai? If you ever think of making UC v1.2 then check this out (it can help with truncations for the tokens limit in chat history)
That said, I can’t deny I would love to outsource certain tasks that wind up taking hours of my life but that would only take AI a minute.
Here is what I want someone to invent: a quote verification app, algorithm, whatever. Basically, I want to be able to feed it a quote, and it returns the original primary source if available or documentation proving it is spurious.
I spend soooo many hours trying to verify quotes, and it seems like half of them turn out to be bogus. I only trust ones where I can see the original sources with my own eyes, if possible, hence my ordering of rare editions of multiple books just to check Aldous Huxley quotes as I share in this piece:
That said, if I were to delegate this work to AI, I would miss out on some absolutely fascinating rabbit-holes, several of which I documented just recently on Notes:
You tried https://quoteinvestigator.com/search/ already, I guess. ChatGPT is useless in verifying research and any app built on current LLM models will be unreliable.
Thank you, Alexander, and yes, Quote Investigator is helpful for those quotes they've happened to investigate, but for the rest, I'm on my own. Archive.org, government congressional records, and Everand have come in quite handy.
This reminds me of the venetian President Luca Zaia who read a poem that seemed particularly fitting for the first COVID-19 lockdown. He said that the poem was written 2000 years ago by Greek poet Heracleion of Gela, but was actually a prank pulled by a computer scientist from Palermo.
Thanks for open-sourcing this Chris. It's such a fantastic tool already and things are just getting started. I know there's a lot of trepidation about AI, much of it for completely valid reasons but I worry that many people don't fully understand the basics and their dogmatic stance against it will prevent them from enjoying the advantages that early adopters have. The power of having an assistant is a force-multiplier.
Alas, some of the people that have vowed to murder me on the public record, have access to state data collection services. And Telegram is part of that apparatus.
~
After an extensive review of the comms software out there, I have decided that private conversations are not possible at this time.
I love how everyone seems to find their own best use case for ChatGPT and other LLMs. Personally I've found ChatGPT to be the most helpful in caring for my house plants and garden. Search engines are terrible at getting you concise information to simple questions like "how much water does this plant like?".
Interesting how you're impatient for something you can ask to "make this whole business". Regret having to bother with your lowly flesh employees? Substack not happening fast enough for your likes? What a dysmally depressing way to see the world. You'll never be satisfied at that rate.
HI Chris, I found the UC today and just finished having a blast with it.
Here's the result: https://astraldwellers.com/
At times I was limited by the amount of tokens. Then I discovered that by deleting some of the .thread content I could unblock it.
I hope you enjoy the game, which is, kind of a child of UC.
This things are great. Power, right? Sort of curated power, instead of raw.
Cheers!
Woah, cool!!
Heard about prompt poet by character.ai? If you ever think of making UC v1.2 then check this out (it can help with truncations for the tokens limit in chat history)
https://research.character.ai/prompt-design-at-character-ai/
On it
Super-cool, Chris!
I have mixed feelings about using AI because, as Emma Sron writes in the following piece, it’s feeding the beast:
• https://emmasron.substack.com/p/artificial-intelligence-ai-the-bigger
That said, I can’t deny I would love to outsource certain tasks that wind up taking hours of my life but that would only take AI a minute.
Here is what I want someone to invent: a quote verification app, algorithm, whatever. Basically, I want to be able to feed it a quote, and it returns the original primary source if available or documentation proving it is spurious.
I spend soooo many hours trying to verify quotes, and it seems like half of them turn out to be bogus. I only trust ones where I can see the original sources with my own eyes, if possible, hence my ordering of rare editions of multiple books just to check Aldous Huxley quotes as I share in this piece:
• https://margaretannaalice.substack.com/p/consequential-quotes-aldous-huxleys
That said, if I were to delegate this work to AI, I would miss out on some absolutely fascinating rabbit-holes, several of which I documented just recently on Notes:
• https://substack.com/@margaretannaalice/note/c-52757995
• https://substack.com/@margaretannaalice/note/c-52778058
I guess I’ll just make peace with manually checking these myself as half the fun is in the journey.
You tried https://quoteinvestigator.com/search/ already, I guess. ChatGPT is useless in verifying research and any app built on current LLM models will be unreliable.
Thank you, Alexander, and yes, Quote Investigator is helpful for those quotes they've happened to investigate, but for the rest, I'm on my own. Archive.org, government congressional records, and Everand have come in quite handy.
This reminds me of the venetian President Luca Zaia who read a poem that seemed particularly fitting for the first COVID-19 lockdown. He said that the poem was written 2000 years ago by Greek poet Heracleion of Gela, but was actually a prank pulled by a computer scientist from Palermo.
Thanks for open-sourcing this Chris. It's such a fantastic tool already and things are just getting started. I know there's a lot of trepidation about AI, much of it for completely valid reasons but I worry that many people don't fully understand the basics and their dogmatic stance against it will prevent them from enjoying the advantages that early adopters have. The power of having an assistant is a force-multiplier.
I published a piece this past week to show how I use LLMs on a daily basis for my creative work and also in my day job. https://www.catchrelease.net/p/ai-denial-its-time-to-get-over-yourself
Thanks for making open source tools! This is what helps make great collaboration possible
If only I understood code … ~ AI Curious
You should have called it _David Deutsch_! Looks super cool :)
This is amazing. I basically built the same terminal app a few months later without having seen yours! https://jamesgrugett.com/p/announcing-manicode-v0
It's taken a while to get it perform file editing at an acceptable level. (It helps that I can use the latest versions of GPT & Claude!)
Nice!!
Yeah it’s funny how hard the editing thing is. I started with gpt4 vanilla and it couldn’t understand line numbers to save its life
Hey Chris,
We are in the same boat of "Where is my programmer assistant!"
Glad to see you jumping out ahead of the pack.
I am currently working on getting Grok-1 going locally...not sure if that is going to work out.
But Somebody out there is converting grok to 4 bit, then it will certainly run locally.
I assume you have played around with ollama.
If not, it's a delight for trying out different models.
Alas, some of the people that have vowed to murder me on the public record, have access to state data collection services. And Telegram is part of that apparatus.
~
After an extensive review of the comms software out there, I have decided that private conversations are not possible at this time.
I am working on fixing that.
I love how everyone seems to find their own best use case for ChatGPT and other LLMs. Personally I've found ChatGPT to be the most helpful in caring for my house plants and garden. Search engines are terrible at getting you concise information to simple questions like "how much water does this plant like?".
Going to read this tomorrow with breakfast, open source in the title got my attention. Open Source FTW 🔥
Interesting how you're impatient for something you can ask to "make this whole business". Regret having to bother with your lowly flesh employees? Substack not happening fast enough for your likes? What a dysmally depressing way to see the world. You'll never be satisfied at that rate.
Cool!
If it starts to suggest we implant chips into monkeys in order to get to the stars then I’d shut it down quick.