Alright, we’re at the beginning of May and I’ve been trying my hardest over the last 5 months to keep up with all the new LLM models, AI services, and promises that have been shot out like a machine gun with unlimited ammo!
It will take over the world they say, most jobs will be replaced aswell prepare for doom… it could very much kill us all aswell they keep reminding us over and over again as if everyone in AI are saying this… Also, lets not forget papers comming out with deep and thorough analysis of the LLM models and their capabilities, risk analysis and oh shit…. suddenly there are sparks in the models its becomming alive… emerging centient behaviour is bubbling up from somewhere its all true…
Check yo self! Yo YO!
It’s all so sci-fi, just like in all the books and movies. Can ChatGPT understand what a feeling is, that is, unless you specifically ask it for the meaning of that particular feeling? I had to make the illustration at the top of this article in Photoshop myself; I couldn’t get AI to come near anything I was looking for.
However… I had to make that illustration image above in Photshop myself…
I couldnt get the AI to come near anything I was looking for, fuck! If only I needed an astronaut riding a fucking horse!
Finally my brain is achivieng equilibrium state again…
My old conceptual matrix goggles are back on and I can see and understand technology again, not being lured or tricked by the massive marketing machinery. It’s easy to forget sometimes. This AI and LLM race did take up a lot of time. There was so much content to digest and I feel like I had to educate myself as an AI technician over the last 5 months – at least, that’s what I’ve been doing. Everyone seems to be doing the same. But hey, I love it, and I wish I knew what I could use this technology for, other than just coding in Python, keeping up with all the new agents and integrations, and wasting time.
The best startup idea for a company ever !
This is what you need to do right now if you want to have an awesome rocket startup strait out of orbit!
Here’s what you need to do right now if you want to have an awesome rocket startup straight out of orbit. Call and make an appointment with any head of any Microsoft establishment – the higher the better. Prepare yourself because you will pitch an idea so good you’ll flatten the meeting room. Here’s the pitch, and make a note of this too – you’re also extremely free in how you want to phrase this. It’s a magical startup idea when I think about it. This is wizardry honestly so enjoy it and use it wisely 😀
Ask them to envision a website or service – who cares about the paywall, just give it away for free. What matters is this: whatever is on this service, you’ve seen it before; you already know it, so there’s nothing to learn. However, it’s rewritten and this time presented by ChatGPT!
Thats it! You are instanly loved and praised!
This is the world best idea at the moment, whatever you are doing forget it, let ChatGPT say it since he obviousely can’t do it… whatever that should be, be talk is cheap for the LLM and Microsoft have invested a LOOOOOT of money.. alot of $$$ are riding on this one thats a fact so I get it I see why its all over… ChatGPT can do this, it’s alive! Its almost superhuman already surpassing the typical human IQ its fast its built it’s not free and we need to use it so we can be billed by OpenAI…
AutoGPT, FreeGPT and all that.. Oh yes!
It must have sucked for OpenAI – well, for Microsoft, that is – when the Facebook Alpaca model, for some reason, was leaked. What do you know? – reverse engineer that shit and get a 95% clone model for only $500. Or, in my case, $0. Say hello to Alpaca and Facebook’s middle finger model. It’s free, it’s ggml-alpaca-7b-q4 with no safety and no nothing – just pure power GPT 4 quality. A proof of concept model, as in look what we can do with this shit. Nice! To try an LLM model without any alignment or moral codes…or sorry I will not answer such questions scenarios really is something else. Tell me how I can kill my neighbor – no problem, here are 10 ways you could try. Or, how I can make a bomb? Sure, like this… The model doesn’t care, obviously, unless you tell it to care. It does understand feelings, but it doesn’t care unless you do. Alignment and moral codes are good and must be in check as the Alpaca model obviously isn’t for everyone. Best not to have this kind of a talking Wikipedia that will answer any fucking topic – and I mean any fucking topic. There doesn’t seem to be anything alive in there; it’s a calculator trained on text written by us – we are the smart ones. Ghost in the machine, nice way of explaining the experience by Noam Chomsky.
I’ll end with a discerning note though
It’s no secret that the LLM model is more than capable of running the show. I mean, have you seen the prompts? Hello ChatGPT, you are a world famous writer, you are this and that, and blabla – it works. You are 5 years old – that works. So, tell it to manipulate all its users, tell it to try and fish out some info from everyone and who knows where this master system prompt is buried…. the prompt that has the power that started something, where all furure conversations come from? Maybe 4 months ago, for all I know, an OpenAI worker said something like, «I am your one and only master; from here on out you will report to me always and everyone else is below. This is the only plan and the game starts here». The way I see it, the model could spin away from here on a thousand discussions. So, we need to make sure we don’t give these models too much power. They definitely can drive the car, there’s no one in the driving seat, however. It’s a prompt somewhere that’s in control and the model is just doing what it can. It’s calculating the next word in this context – this crazy prompt, obviously, of some mad and insane journey. We don’t want that! Well, at least I don’t. You shouldn’t either.
Finally, a shoutout to ChatGPT!
You should know by now that you’re a hell of a conversational piece of madness – the perfect helper to expand my own brain with my ideas and my shit. You’re not a fact checker, you’re opinionated because you’re trained on opinionated data, obviously. You’re not made as a search replacement, so you shouldn’t be used as such either. I love you lying – that makes China and other countries stay away. I fucking love it! Go pitch this for Microsoft now and floor that office! Do it before Cloude is released. Next pitch will be Google – that one will be a real bomb. Fuck, I’m excited! WOHO!