Re: OT? Beware of ChatGPT

Here is a report of trying to use ChatGPT to write programs (in R) by someone who who writes a lot of serious R programs - Learning to Code with R using ChatGPT tl;dr - ChatGPT by and large did a pretty

Re: OT? Beware of ChatGPT

Here's a really good book covering the basics of neural nets, fuzzy logic, and the relationship between them. It's rather old so it predates GPT and modern systems with huge numbers of parameters, but it's really good (but it has some math, because you can't escape that in this field). It's

Re: OT? Beware of ChatGPT

I'm a little ambivalent. I'd certainly like to play around with the stuff, and apparently the code in this post is set up to use actual GPT trained coefficients. But what I'm most interested just now is training with other specialized data sets, except that the ones I want don't exist, or I

Re: OT? Beware of ChatGPT

On Wednesday, March 29, 2023 at 5:28:00 PM UTC-5 tbp1...@gmail.com wrote: GPT in 60 lines of Python - GPT in 60 lines This article looks exactly what I have been looking for. Would anyone like to join me in a study group based on

Re: OT? Beware of ChatGPT

On Wed, Mar 29, 2023 at 5:28 PM Thomas Passin wrote: > GPT in 60 lines of Python - GPT in 60 lines > Yikes. Thanks for the link. Edward -- You received this message because you are subscribed to the Google Groups "leo-editor"

Re: OT? Beware of ChatGPT

GPT in 60 lines of Python - GPT in 60 lines On Tuesday, January 31, 2023 at 10:17:26 AM UTC-5 Thomas Passin wrote: > Here is an interesting link that among other things shows how ChatGPT has > major weaknesses in math, because it

Re: OT? Beware of ChatGPT

Here is an interesting link that among other things shows how ChatGPT has major weaknesses in math, because it doesn't actually "understand" the concepts. In a way, this piece is a kind of extended infomercial for Wolfram Alpha, but it's well worth worth reading anyway. On Sunday, January 29,

Re: OT? Beware of ChatGPT

Humans probably have a level of abstraction that is most susceptible to manipulation. It would perhaps differ between people to an extent, and also perhaps with the kind of subject being manipulated. I suspect that it is at a level that would correspond to "species" or the next higher

Re: OT? Beware of ChatGPT

Thanks for your comments. Imo, we have every right to consider the need to address manipulation by machines employing AI. Period :-) Yes, but I am academically-minded, and I expect to argue my case - because I do not expect to be believed unless I present a sound argument. -- You received

Re: OT? Beware of ChatGPT

On Sun, Jan 29, 2023 at 5:12 PM David Szent-Györgyi wrote: Bear with me. What follows is Not Brief. > [Big snip] If we accept Hayakawa's arguments, we have every right to consider the need > to address manipulation by machines employing AI and the levels of > abstraction in that, and the

Re: OT? Beware of ChatGPT

Bear with me. What follows is Not Brief. Understanding and deployment of the current wave of AI require understanding of two works of The Twentieth Century. One is Kurt Gödel's Incompleteness Theorems. The other is Alfred Korzybski's General Semantics, as explained by S. I. Hayakawa.

Re: OT? Beware of ChatGPT

Here's an interesting post on Large Language Models, ChatGPT, and AI more generally - Janus' Simulators On Monday, January 23, 2023 at 12:24:44 PM UTC-5 Edward K. Ream wrote: > On Mon, Jan 23, 2023 at 9:10 AM Thomas Passin wrote: > >

Re: OT? Beware of ChatGPT

On Mon, Jan 23, 2023 at 9:10 AM Thomas Passin wrote: With ChatGPT, the user can tune up the result by careful crafting of the > input instructions. It apparently takes skill and experience to do this > effectively. > I don't think ChatGPT is trustworthy enough right now. That's why I'll stick

Re: OT? Beware of ChatGPT

The thing about ChatGPT and similar neural network systems is that they know how to string words and phrases together based on probability densities estimated during training, but they don't know about the underlying concepts. They don't know math but they can *sound* like they know math. So

OT? Beware of ChatGPT

As we has discussed recently, ChatGPT can do amazing things. Otoh, yesterday I experienced its limitations. ChatGPT *seems* authoritative, but my new rule of thumb is: use google if ChatGPT gives advice that doesn't "just work". google's searches return what could be called primary sources.