mcbuilder

8 ms·

121.
by mcbuilder 3y ago
Writing in a language that I don't know that well, but I know enough. Okay, my javascript isn't the strongest, so I'd prolly have to spend 30-45 minutes just coming up to speed again on my basic AJAX and modern syntax, or BAM
122.
by mcbuilder 3y ago
I copy pasted the attention is all you need paper into ChatGPT4 and gave it the prompt "Explain like I'm 5 years old". The Transformer is a new type of computer program that helps translate languages and understand sentences
123.
by mcbuilder 3y ago
It's so easy to fall for it as well, even as someone who has been using LLMs since they came out. Just the other day, I was using ChatGPT on my phone, when I found myself asking whether it was GPT-3.5 or GPT-4. Didn't realize I wa
124.
by mcbuilder 3y ago
These models can run FP16, with LLM quantization going down to Int8 and beyond.
125.
by mcbuilder 3y ago
Am I the only person who hates the idea of "alignment"?
126.
by mcbuilder 3y ago
That's the sort of blue sky research I'm glad exists.
127.
by mcbuilder 3y ago
But yet it can understand a json data schema from example and write javascript to interact with a library that I fed it and asked it to understand. Yes, I know its limitations, but it can also surprise me.
128.
by mcbuilder 3y ago
Looks like a quick blog post by a PI outlining 3 of their lab's projects and giving a quick overview of their research directions. I wouldn't take it as an informational article in the sense you're looking for. It's more
129.
by mcbuilder 4y ago
Is it strange that I think it's more likely that LLMs generalize and synthesize new things out of old patterns instead of just copying a singular example.
130.
by mcbuilder 4y ago
Classic doubled edged sword, perhaps the one stable thing about so called technological revolutions. Yay industrialization, boo child workers. Yay, social media, boo increased teenage depression, Yay AI, boo AI.
131.
by mcbuilder 4y ago
I've had a macro for that in EMACS for the past 30 years.
132.
by mcbuilder 4y ago
I don't know about about this model, but usually with these ML models you download the static weights, but nothing is stopping you from fine tuning them to your needs/new information. It's not automatic, would require some ML
133.
by mcbuilder 4y ago
Yeah I grew up doing Transcendental Meditation, the instruction was that any effort was too much effort. It took me so long to grasp this concept, but eventually it clicked. It's a hard concept to convey, because lots of things can com
134.
by mcbuilder 4y ago
Sometimes I wonder why I keep a boxes full of a good dozen late IBM era Thinkpads, and I remember thinks like this exist and hopefully will continue into the future with modern tech like OLED, power efficient chips. Maybe if I even just use
135.
by mcbuilder 4y ago
And at 9th grade it's a free for all? Looking back the high school years would have been the worst for my own self esteem, angst, etc. Glad none of it was recorded. :)
136.
by mcbuilder 4y ago
Yeah, I remember how things were before yuge public LLM (don't you? it was like a few months ago), and a PR official would have roasted for picking a Hallmark card at anytime in my lifetime.
137.
by mcbuilder 4y ago
Ha Ha, great test. I modified this into a prompt and now have a ChatGPT prompt: ``` Task: Was this written by ChatGPT? And Why? Test Phrase: "Yes, the message you provided is likely to be spam. The message contains several red flags in
138.
by mcbuilder 4y ago
I think it seems likely any thing similar to a blog farm you describe would also get detected by the AI. Maybe we will just develop AI bullshit filters (well embeddings) just like I can download a porn blacklist or a spam filter for my emai
139.
by mcbuilder 4y ago
I get that with my kids. I'm not an iPhone user, so I don't know the app store. But my very young kid will know how to open the app store on my wife's iPhone. With Android I type "open source" or use FDroid if I don
140.
by mcbuilder 4y ago
Great, not only is ChatGPT better at rhyming than me, it also sounds more sophisticated while doing it. LLMs fail to pass for sophomoric when viewed by an expert, but to the mediocre it exposes one's shortcomings. Maybe because they so
141.
by mcbuilder 4y ago
This is incredible, think of the applications! Running a terminal inside a WASM VM inside a browser running from your VT100 terminal is now possible. I am literally floored, and I've been waiting for this moment for years. Thank you fo
142.
by mcbuilder 4y ago
Someone has to pay to run the AI though. With Bitcoin it's a bit different, you mine and you get money directly. Even if you have a AI burning GPU cycles all day generating content, you still gotta turn it into cash.
143.
by mcbuilder 4y ago
It's not that different than naming a Protein by a Greek god or goddess, is it?
144.
by mcbuilder 4y ago
I don't even think it extends to this, it's simply because it's automated. I have no talent in the area, but I know that artists can copy one another's styles. You see it in talented art student master copies, and hell I
145.
by mcbuilder 4y ago
It works, I was watching some "normie" youtube creator using Chat-GPT in a face-off with him over some domain specific question and answers. He was new to the LLM thing you could tell, but said how it was an open source right in t
146.
by mcbuilder 4y ago
It's really crazy the lengths people go to "filter" these models and limit their output, and of course soon these filter will be a another level of "AI" (see Pathways or any mixture of experts, maybe add some contex
147.
by mcbuilder 4y ago
Heretic!
148.
by mcbuilder 4y ago
Yeah, easily $1k more if I could have that and perfect Linux support.
149.
by mcbuilder 4y ago
This is "Thanks" by some poor ops engineer trying to do her job. You'd have her fired if the company lost a lawsuit all to prove some bullshit "ideal". The two things are completely different, for one a newspaper ha
150.
by mcbuilder 4y ago
Well if it used to cost you $1 for 1hr at 1x speed, now it will take you 10hr at 0.1x speed, and if my math checks out $1. You need to shrink the model.

More ›