Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

ChatGPT Plugins

openai.com

Post reply on HN
Print view View original

211–220 of 1001 posts

Re: ChatGPT Plugins

#211

Post by kokanee »

sipjca wrote: post #188
celestialcheese wrote: post #39

This is a big deal for openai. Been working with homegrown toolkits and langchain, the open source version of this, for a number of months and the ability to call out to vectorstores, serpapis, etc, and chaining together generations and data-retrieval really unlocks the power of the LLMs. That being said, I'd never build anything dependent on these plugins. OpenAI and their models rule the day today, but who knows wh…

i think local ai systems are inevitable. we continue to get better compute, and even today we can run more primitive models directly on an iPhone. the future exists in low power compute running models of the caliber of gpt-4 inferring in near-realtime

The technical capability is inevitable, but remember that people hate doing things themselves, and have proven time and time again that they will overlook all kinds of nasty behavior in exchange for consumer grade experiences. The marketplace loves centralization.

Re: ChatGPT Plugins

#212

Post by yunyu »

qbasic_forever wrote: post #94
londons_explore wrote: post #79

Earlier quoted context omitted.

See the link in my post. It asks you to run the tool. You run the tool and tell it the result... And then it uses the result of the tool to decide to reply to the user. The link talks about tools that 'lie' - ie. a calculator which deliberately tries to trick GPT-4 into giving the wrong answer. It turns out that GPT-4 only trusts the tools to a certain extent - if the answer the tool gives is too unbelievable, then G…

It's always giving a hallucinated answer. GPT doesn't 'run' anything. It sees an input string of text asking for the result of fibonacci(100) and finds from its immense training set a response that's closely related to training data that had the result of fibonacci(100) (an extremely common programming exercise with results all over the internet and presumably its training data). Again, GPT is not running a tool or a…

There's nothing stopping you from identifying the code, running it, and passing the output back into the context window.

Re: ChatGPT Plugins

#214

Post by davidkunz »

Plugins I would like to see:

- Compiler/parser for programming languages (to see if code compiles)

- Read and write access to a given directory on the file system (to automatically change a code base)

- Access to given tools, to be invoked in that directory (cargo test, npm test, ...)

Then I could just say what I want, lean back and have a functioning program in the end.

Re: ChatGPT Plugins

#215

Post by HarHarVeryFunny »

akavi wrote: post #55

I've got to wonder, how does a second player in the LLM space even get on the board? Like, this feels a lot like when the iPhone jumped out to grab the lion share of mobile. But the switching costs was much smaller (end users could just go out and buy an Android phone), and network effects much weaker (synergy with iTunes and the famous blue bubbles... and that's about it). Here it feels like a lot of the value is em…

Google have really been caught with their pants down here.

Remember that OpenAI was created specifically to stave off the threat of AI monopolization by Google (or anyone else - but at the time Google).

DeepMind have done some interesting stuff with Go, Protein folding etc, but nothing really commercial, nor addressing their reason d'etre of AGI.

Google's just-released ChatGPT competitor, Bard, seems surprisingly weak, and meantime OpenAI are just widening their lead. Seems like a case of the small nimble startup running circles around the big corporate behemoth.

Re: ChatGPT Plugins

#216

Post by sfink »

dougmwne wrote: post #30

I would be interested to play with a long term memory plugin. It could be a note-taking system that would summarize prior conversations and pull their context into the current conversation through topic searches. This would enable the model to have a blurry long term memory outside of the current context. I played with some prompts and GTP-4 seems to have no problem reading and writing to a simulated long term memory…

"Grandpa, we know you've been really bothered by your memory loss and you're happy that you've come up with a way to fix it.

"But we really think you need to get this thing under better control.

"Your granddaughter's name is indeed Alice, but she's only 3: she is not running a pedophile ring out of a pizza parlor. Your neighbor's house burned down because of an electrical short, it was not zapped with a Jewish space laser.

"Now switch that thing off and go do something about the line of trucks outside that are trying to deliver the 3129833 pounds of flour you ordered for your halved pancake recipe."

Re: ChatGPT Plugins

#217

Post by visarga »

endisneigh wrote: post #159

I see a lot of positive sentiment and hype, but ultimately unless they own the phone ecosystem they will lose in the end, imho. In a year Apple and Google will trivially create something equivalent. Those who control the full stack (hardware, software and ecosystem) will be the true winners.

True, this will not only be replicated by Google, Apple, Amazon and Facebook, but also by open-source. OpenAI has a short window of exclusivity. Nobody can afford to wait now, after reading the Sparks of Artificial General Intelligence paper I am convinced it is proto-AGI. Just read the math section, coding and tool use. I've read thousands of papers and never have I seen one like this.

https://arxiv.org/pdf/2303.12712.pdf

Re: ChatGPT Plugins

#218

Post by taf2 »

celestialcheese wrote: post #39

This is a big deal for openai. Been working with homegrown toolkits and langchain, the open source version of this, for a number of months and the ability to call out to vectorstores, serpapis, etc, and chaining together generations and data-retrieval really unlocks the power of the LLMs. That being said, I'd never build anything dependent on these plugins. OpenAI and their models rule the day today, but who knows wh…

I think you're right... but ChatGPT is just so damn good and the price is 0.002 per 1k tokens is very easy to consume... It is a big risk that they can't maintain compatibility or that they fail or a competitor emerges that provides a more economical or sufficiently better solution. They might also just becomes so unreliable because their selected price isn't sustainable (too good to last)... For now though they're too good and too cheap to ignore...

Re: ChatGPT Plugins

#219

Post by radus »

davidkunz wrote: post #214

Plugins I would like to see: - Compiler/parser for programming languages (to see if code compiles) - Read and write access to a given directory on the file system (to automatically change a code base) - Access to given tools, to be invoked in that directory (cargo test, npm test, ...) Then I could just say what I want, lean back and have a functioning program in the end.

I'm sure this type of integration will happen, but... isn't this exactly how AGI would "escape"?

Re: ChatGPT Plugins

#220

Post by 93po »

mikeknoop wrote: post #17

(Zapier cofounder) Super excited for this. Tool use for LLMs goes way beyond just search. Zapier is a launch partner here -- you can access any of the 5k+ apps / 20k+ actions on Zapier directly from within ChatGPT. We are eager to see how folks leverage this composability. Some new example capabilities are retrieving data from any app, draft and send messages/emails, and complex multi step reasoning like look up data…

I saw a startup recently that's working to automate interactions with applications that are either not web apps (in which case you'd run a local instance of it) or a web app that doesn't provide an API to do certain (or any) actions. Is this something Zapier is looking at, too? It would really expand what's possible with the OpenAI integration and save people a tremendous amount of time to not be forced to jump through hoops interacting with often crappy software.
Post reply on HN