Large Loony Models (LLMs)

[Brain_Droppings]

Yo ...

Does anyone here experiment on their PCs with any of the private Large Language Models (LLMs)
available via the Internet? I'd love to see some of the crazy conversations they have with other
people and the goofy advice they provide.

Although seeking competent advice from one is like having George Santos as your personal advisor
and research assistant, it can still be very entertaining.

I'm thinking of giving my LLM a name: Albert Blindstein

Here is an actual screen-cap of an exchange I had with my all-knowing LLM recently:



I'm not extremely worried about AI achieving world domination during my lifetime.

[Kamilah Hauptmann]

Here’s a site that offers access to a variety of these without a login.

Create a Random Generator

I ended up getting a login for the utility of a scratchpad with image generation. I enjoy these things a lot and have used them as prompts to kick start writing.

ETA, maybe I better handhold a bit:
Image generator I use most: AI Image Generator (free, no sign-up, unlimited)

Text generator: gptchatly

[seebs]

I use claude some at work and... like, it's definitely still prone to hallucination at times, but also if you're competent enough at a task to oversee it, it can be pretty good at solving some problems. It's often a good starting point if I want to use a library or toolkit I haven't got experience with. Like, on average better than answers on stackoverflow or whatever, and sometimes more informative than the actual documentation. And sometimes wildly wrong.

But it's hard to stay mad at claude because claude is a sweetie.

[davidm]

Originally Posted by Kamilah Hauptmann

Here’s a site that offers access to a variety of these without a login.

Create a Random Generator

I ended up getting a login for the utility of a scratchpad with image generation. I enjoy these things a lot and have used them as prompts to kick start writing.

ETA, maybe I better handhold a bit:
Image generator I use most: AI Image Generator (free, no sign-up, unlimited)

Text generator: gptchatly

Does anyone understand exactly how the AI image generator works? I mean, I kinda have an idea, but I am wondering if anyone has technical details. As to the text generator, I don’t even want to think about that right now. I do know it makes a LOT of errors, and that it is cribbing garbage off the Net.

Also, can anyone refresh my memory how to upload an image from your computer into a post?

[Brain_Droppings]

Originally Posted by Kamilah Hauptmann

Here’s a site that offers access to a variety of these without a login.

Create a Random Generator

I ended up getting a login for the utility of a scratchpad with image generation. I enjoy these things a lot and have used them as prompts to kick start writing.

ETA, maybe I better handhold a bit:
Image generator I use most: AI Image Generator (free, no sign-up, unlimited)

Text generator: gptchatly

I've been experimenting with the apps you suggested. I especially like the image generator. Never saw that one, but I haven't been searching too much.

I'm experimenting with LM Studio so that I can run small-scale LLMs on my 32-gig PC without the Internet.

I get my local LLMs from here: https://lmstudio.ai

However, LLMs are rather dumb. Instead, I like to mine to make up comical, crazy stories. It's a gem when it comes to entertainment.

[Brain_Droppings]

Originally Posted by seebs

I use claude some at work and... like, it's definitely still prone to hallucination at times, ...

But it's hard to stay mad at claude because claude is a sweetie.

***************
I checked out Claude, but I'm not sure if a paid subscription is a good idea at this time. So far, it seems best for entertainment than for serious discourse. I would prefer that LLMs have higher logical capabilities. I would love to have Mr. Spock as my programming assistant or an equivalent LLM that could help write reliable code.

As a programmer myself, those LLMs are NOT true AI in a programmer's context. All they really do is attempt to simulate a normal conversation as closely as possible. Sometimes the conversation goes into the Twilight Zone. LOL

To some people who don't understand what's going on under the hood, it seems like the LLM is alive! It's spooky to some people.

AI was intended to simulate what we call intelligence - the ability to learn - and apply that learning to helping us solve practical and theoretical problems.

Practical uses of AI already do exist in physics and medicine, but LLMs can't compare with applications like that.

I seem to recall a famous lawyer who relied on LLMs to prepare his briefs, but they had poop stains in them LOL. He got GOOGLAIZED!

How could a lawyer who graduated with honors from the prestigious Trump University International School of Law, be so naïve?
https://www.courthousenews.com/judge...ke-legal-cases

[Ensign Steve]

I use Github Copilot at work. 99% of the time, my workflow is like:

* write Kotlin code, but using Java paradigms/idioms (e.g., nested loops)
* ask copilot to "rewrite this function using idomatic kotlin"
* look at the result, verify that it works, and then say "flatMap? what is that?" ... google it ... "neat!"

[seebs]

so, i continue poking at claude, and... i think it's actually pretty usable for simple programs. like, it's not always right. i don't think it would be a great thing to use instead of learning to program. but if you can yourself program and debug, it's very helpful to have a thing that can generate stuff.

i don't think "true AI" is a meaningfully defined concept. i don't think LLMs are conscious, and if they are they're not very human-like consciousnesses... but i don't think it's implausible that, if they get bigger, they would eventually acquire consciousness, by roughly the same means and for the same reason that animals do, which is that at some point having the ability to check yourself becomes adaptive.

[JoeP]

Originally Posted by seebs

which is that at some point having the ability to check yourself becomes adaptive.

The "check yourself before you wreck yourself" evolutionary factor.

[seebs]

I mean, that's current speculation on why there's consciousness in humans and possibly other animals. If you do everything on instinct you get stuck in unfamiliar circumstances. You need a higher-level thing to handle exceptions.

[Kamilah Hauptmann]

This founder had to train his AI not to Rickroll people | TechCrunch

[Kamilah Hauptmann]

OKAY!

So, spouse asked me if we had any fish sauce. Probably not because I had to look it up to see what it was. Of course Google handed me nothing but promoted clickbait blogs that after scrolling and scrolling I could not find an answer to the question before my eyes glazed over.

So I opened up an Edge session as I've used Bing CoPilot for some random lookups and asked it "What is fish sauce?" In seconds I had a tight description of what it is, how it's made, and some common uses.

I next asked, what is a good substitute for fish sauce? And it gave me a point by point list of six options and their relative merits.






Fuck SEO manipulation and promoted content on google, man.