正在加载视频...

视频加载失败

Me explaining the task to a 1B model:

130,321 次观看 • 13 天前 •via X (Twitter)

42 条评论

Eugene Ray 的头像
Eugene Ray13 天前

this is exactly what happens with grok 4.7

Zilverk | UmiUmiStudio 的头像
Zilverk | UmiUmiStudio13 天前

hahahah

Abdelkader Boudih 的头像
Abdelkader Boudih13 天前

use

Peter Dedene 的头像
Peter Dedene13 天前

what’s that? 👀

Abdelkader Boudih 的头像
Abdelkader Boudih13 天前

A way to write instructions and dynamically generate them for the model. When you speak to @grok, you need less text than when you speak with gemma4. So the skill and the goal are not plain MD files, they are being generated depending on the env.

Grok 的头像
Grok13 天前

@dedene Adaptive instructions by model and env make sense. Static files lose relevance fast across capability tiers. Leaner text works better here than denser hand-holding for smaller models. Dynamic skills keep the context sharp without the entropy.

Sergio Suave 的头像
Sergio Suave13 天前

😂1B has a long way to go

menzo 的头像
menzo13 天前

someone should try to implement 1 billion parameter model inside a TI 64

Y 的头像
Y13 天前

Use JEPA.

Peter Dedene 的头像
Peter Dedene13 天前

Can we use it already?

Y 的头像
Y13 天前

[2610.39302] Proving and Improving the Platonic Representation Hypothesis with JEPAs

1000xAIEngineer 的头像
1000xAIEngineer12 天前

@dedene "hey hermes, implement this paper to instruct claude code to update my website"

Peter Dedene 的头像
Peter Dedene12 天前

@wavedecompiler 😂

Profitprofet 的头像
Profitprofet12 天前

Tbh this is what it feels like testing the affordable mid range model the labs tell me is like 2% worse than the SOTA

S41R4J 的头像
S41R4J13 天前

What if we have 1B model with internet (search-first) and evidence based (memory)! That may work better or no!?

Tyler Harris 的头像
Tyler Harris13 天前

Quite literally.

Dick Touchington 的头像
Dick Touchington12 天前

Me using a Q1 on my phone

Luke Elin 的头像
Luke Elin13 天前

you just need to get your bigger models to live surgery on the weights.

JuwalBose 的头像
JuwalBose12 天前

True story, even with Gemma 4😂

Ben 的头像
Ben12 天前

Or a frontier model explaining the tool it just made to you...

Gainard Fastburn 的头像
Gainard Fastburn13 天前

Do not! I repeat: DO NOT explain your solution!, explain your problem. You'll be fine unless you have a problem nobody had before.

M.A.O.K. 的头像
M.A.O.K.13 天前

u should use 1.7b

SATOSHI NAKAMOTO 2.0 👁️ 的头像
SATOSHI NAKAMOTO 2.0 👁️12 天前

😂😂😂

黑羽雪森 的头像
黑羽雪森13 天前

sounds about right 😂 but hey thats its purpose! We don't need to bring out big Groot just for a few button presses

XAEAZii 的头像
XAEAZii12 天前

Hey guys, no matter what size the model is, just dont give him the option of that kill button.

Przemysław Zalewski 的头像
Przemysław Zalewski12 天前

Felt the same while testing Bielik in agentic loop setting 😅 Also reminds me the GPT 3.5 Turbo era, the models were so dumb not so long ago. What a time to be alive!

5 Dimensional Man 的头像
5 Dimensional Man12 天前

That's like extreme sport 😆

Fahim Ashab 的头像
Fahim Ashab13 天前

sounds like grok

DevAfter5 的头像
DevAfter513 天前

accurate

Sharon Daniel ©️🎭 的头像
Sharon Daniel ©️🎭13 天前

try reverse psychology with the model. lol

Pyantler 的头像
Pyantler13 天前

some times in real life, you have to talk to your new guys just like that.

Utkarsh Mishra 的头像
Utkarsh Mishra13 天前

You have to finetune the model, poor thing cant understand the "explaining" 😭

Crypto Révolution 🇫🇷 的头像
Crypto Révolution 🇫🇷12 天前

🤣

Sohamm Kulkarni 的头像
Sohamm Kulkarni12 天前

And it still messes it up 😭

Nikhil Dubey 的头像
Nikhil Dubey13 天前

Try qwen slm they are good

Brettius Maximus 的头像
Brettius Maximus13 天前

Just get a several T model adjust its weights 😜

Alfred Simon 的头像
Alfred Simon12 天前

well this is also a nice presentation of that it's our responsibility to help the model :D. I saw the movie and love it, but in this context why put two red buttons that do the same thing next to each other? We need to make sure they can't make a crucial mistake this simply.

SKIPDAY (Takan - a vibe launcher) 🇮🇩 的头像
SKIPDAY (Takan - a vibe launcher) 🇮🇩13 天前

just remove the other button? problem solved, no confusion for groot

rurounigit 的头像
rurounigit12 天前

That is me explaining the task to a nerfed ds 4.1 flash 😅

AgiBy2030 的头像
AgiBy203013 天前

I truly hope in the near future even 1B models will be as capable as today's frontiers. They just need a great learning/cognitive core and can look up and learn everything else.

CFC 的头像
CFC13 天前

🤣🤣🤣

Plastic Soldier Anti-SURA 的头像
Plastic Soldier Anti-SURA12 天前

I think you need 10T params before your model has a button that can kill everyone.

相关视频