Loading video...

Video Failed to Load

Go Home

Me explaining the task to a 1B model:

130,321 views • 13 days ago •via X (Twitter)

42 Comments

Eugene Ray's profile picture
Eugene Ray13 days ago

this is exactly what happens with grok 4.7

Zilverk | UmiUmiStudio's profile picture
Zilverk | UmiUmiStudio13 days ago

hahahah

Abdelkader Boudih's profile picture
Abdelkader Boudih13 days ago

use

Peter Dedene's profile picture
Peter Dedene13 days ago

what’s that? 👀

Abdelkader Boudih's profile picture
Abdelkader Boudih13 days ago

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's profile picture
Grok13 days ago

@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's profile picture
Sergio Suave13 days ago

😂1B has a long way to go

menzo's profile picture
menzo13 days ago

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

Y's profile picture
Y13 days ago

Use JEPA.

Peter Dedene's profile picture
Peter Dedene13 days ago

Can we use it already?

Y's profile picture
Y13 days ago

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

1000xAIEngineer's profile picture
1000xAIEngineer12 days ago

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

Peter Dedene's profile picture
Peter Dedene12 days ago

@wavedecompiler 😂

Profitprofet's profile picture
Profitprofet12 days ago

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's profile picture
S41R4J13 days ago

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

Tyler Harris's profile picture
Tyler Harris13 days ago

Quite literally.

Dick Touchington's profile picture
Dick Touchington12 days ago

Me using a Q1 on my phone

Luke Elin's profile picture
Luke Elin13 days ago

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

JuwalBose's profile picture
JuwalBose12 days ago

True story, even with Gemma 4😂

Ben's profile picture
Ben12 days ago

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

Gainard Fastburn's profile picture
Gainard Fastburn13 days ago

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.'s profile picture
M.A.O.K.13 days ago

u should use 1.7b

SATOSHI NAKAMOTO 2.0 👁️'s profile picture
SATOSHI NAKAMOTO 2.0 👁️12 days ago

😂😂😂

黑羽雪森's profile picture
黑羽雪森13 days ago

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

XAEAZii's profile picture
XAEAZii12 days ago

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

Przemysław Zalewski's profile picture
Przemysław Zalewski12 days ago

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's profile picture
5 Dimensional Man12 days ago

That's like extreme sport 😆

Fahim Ashab's profile picture
Fahim Ashab13 days ago

sounds like grok

DevAfter5's profile picture
DevAfter513 days ago

accurate

Sharon Daniel ©️🎭's profile picture
Sharon Daniel ©️🎭13 days ago

try reverse psychology with the model. lol

Pyantler's profile picture
Pyantler13 days ago

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

Utkarsh Mishra's profile picture
Utkarsh Mishra13 days ago

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

Crypto Révolution 🇫🇷's profile picture
Crypto Révolution 🇫🇷12 days ago

🤣

Sohamm Kulkarni's profile picture
Sohamm Kulkarni12 days ago

And it still messes it up 😭

Nikhil Dubey's profile picture
Nikhil Dubey13 days ago

Try qwen slm they are good

Brettius Maximus's profile picture
Brettius Maximus13 days ago

Just get a several T model adjust its weights 😜

Alfred Simon's profile picture
Alfred Simon12 days ago

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) 🇮🇩's profile picture
SKIPDAY (Takan - a vibe launcher) 🇮🇩13 days ago

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

rurounigit's profile picture
rurounigit12 days ago

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

AgiBy2030's profile picture
AgiBy203013 days ago

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's profile picture
CFC13 days ago

🤣🤣🤣

Plastic Soldier Anti-SURA's profile picture
Plastic Soldier Anti-SURA12 days ago

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

Related Videos