Replica apparently didn't use LLMs in 2018 though, so not really. Transformers were made in 2017 by google and then openai scaled it up but replica was not using that in 2018. If you mean chat bots in general then those were around before 2000s and Id have used those since I visited a science museum that had one in the 2000s... But that's not really an LLM. Id argue copilot was the first actually useful LLM with broad utility, the rest was trash with niche applications.
Make that 6 years for LLM ( Since it wasn't available, but 8 years with Ai.
When I was mining cryptocurrency in 2017 selling hashpower for machine learning was a 2nd option. Ai was always a thing and has been beating our ass in chess and video games for some time now, but when i learned about machine learning I had to see what it's for, and that's when I got my first chatbot.
Didn't use LLMs though.
Not all Ai are large language models.
.
.
.
In the end it doesn't matter who has the most experience using Ai.
Well it does if someone says that I was the one behind on using AI out of all of us when I was the first one to be using LLMs for real applications, .
The LLM madness started earlier than chatgpt.
GPT-1 being the first LLM made public in 2018.
I recall you complaining about Copilot back in 2025. How shit it was, That's 1 year ago. Right, so since GPT-3 ( which is even worse ) you've been using Ai that's next to useless for coding. Unless you wrote code for machine learning, I don't see using it as something to boast about. At best using Ai is like asking someone to do something for us.
It's no longer important to know how to prompt properly or use python to issue prompts, as the most powerful Ai can be instructed in plain English on what to do and how to do it, and it will know what you mean in ways some people don't seem to convey. Just speak/write, it will confirm your request and let it rip. Thing knows how to cook.
Its worth knowing how it works. I see too many people glazing AI without understanding the fundamentals. People dont even know how transformers work or the foundational technology. They dont understand how the AI predicts what its supposed to output and what it cant do. Once you understand the technology behind it, its easy to see the limitations. im the only one here capable of understanding the technology and Ive used LLMs when nobody here knew anything about them. I also know what the difference is between LLMs and general chat bots.
So, arguably I am the most proficient in AI here and always were. Its just what it is.
That doesn't make you a star or a better than anyone who uses modern Ai.
I'd argue when I apply Ai to what I do ( which is different than what you use it for ) my output is finer. At present I'm not just having Ai do things for me, I have it doing things with me. I'm a long time Autodesk Maya user, and when I bring Autodesk Flow Studio into the mix, I'm doing more than just directing generative Video, I'm controlling it in ways a director couldn't possibly do.
What you do for a living is, well, I don't know you're not solving the big one are you.
Behind the components not even the developers understand how Ai is even working.
Ai itself is proficient so the user can move forward. I understand the combustion engine, and have taken them apart and put them back together, but that doesn't matter cause it doesn't make me a better driver.
At this point I use multiple Ai tools and the only person in this community, to post more than LLM output.
I dont know what this means.
It means I've shown the most Ai output, and it's not simply using LLM Ai.
It sounds mean but ChatGPT kinda makes me cringe though it's very nice, and I mean very very nice, it's more like a brand for people who don't make it very far.
I see most of everyone this way given that they dont understand what an LLM is.
I see you wanting to merge with it as a form of vanity. Hence how flustered you sound.
Its all well and good. If you dont use Linux and are not building a system together with the AI Id argue youre practically handicapped by the limitations of LLMs.
I've done that and ended up having to use my main PC cause the solo build I scrapped together was too pleb and would only handle 7b models.
At the end of the day when we build a system to run an LLM locally, using a harness like Hermes 2.0 it's still going to choose the public Ai as a flagship brain when we want to get real and code.
This was before SOL came out so the harness would NEVER use GPT as the brain for what I'd do with it.
Simply put, there is no local Ai that can outclass paid public Ai. Doesn't matter if you built your own data center with state of the art hardware, it's not happening. If you're slightly above normal you'll find RTX XX90 GPUs are lacking sufficient VRAM and cannot content with what data centers have to offer.
In my experience with OpenClaw and Hermes, it's simply not as good as it looks. When it comes to Claude Ai, we wouldn't want it to use anything higher than Opus 4.5 or else it will gobble up tokens real quick, it's simply better to use the Claude Ai than to use it in a harness. That in itself kills off using a Harness.
Same shit. I say use Gemini instead, it'll give you free usage all day long plus you're in debt. Nope. I'll have to get her friend to use it first.
When I look into the whole "Ai made me do it and now I'm fucked" scene, I see a bunch of dumbasses using ChatGPT, and sure I look snooty to them, while I'm making apps and websites and collaborating with Ai to make nice games and sell shit. I see people using ChatGPT to fuel drama and output high quality nothing burgers.
What kind of games did you make?
Ive used it to tailor my OS to my liking and applications I dont want to make myself. I automate everything with AI, which means I can focus all of my time on research now. Its nice.
There are 3 on the backburner, one I just started.
The first one is a concept. And will require ComfyUi for the background. This one will be cheap and I never started it. The 2nd one uses sprites made with Ai and myself. The 3rd one, I just started 3 days ago. My team consists of Ox Alpha, Claude, and Gemini. While I'm a Grok fan, that's included too sometimes. Claude is the Captain while Gemini is just really good at inserting itself when I turn to it for free perspectives. It'll say, awesome can I see, Okay you want to see what Claude wrote here. Boom oooo you can do this and this and this, and I'm like good idea you're in.
Ox Alpha is kinda funny to use cause this thing will not remember shit. This one is free to use and it can code. So before I leave it alone I will make it create a "note to self" and show it to the next instance along with everything it's done before carrying on with the session.
My games are not just done in a day. 2 of them are meh. While the new one I think will be a hit. I don't know how long it'll take before I release it. As it is in 3D, it's never done, and I'm the one setting a deadline which is, "however long it takes".
