sc

[sc @space_cap]

@EHuanglu Space time and a explorer.
Albert Einstein would love this!
We traveling as explorer in this world or does the world let us travel? We need to understand the forces.

[sc @space_cap]

@4dimensionapp Start with a simple planetary system this will help to understand spacemechanics it is a bit less chaotic.

[sc @space_cap]

@4dimensionapp In this visuals the time is still mixed with the forces of time. But could be good starting point to get there.
In space scale we have gravitational forces. Like 2d pixel image deep visual., we could get there, each voxel (mass) gets a gravitational vector +light raytracing.

[sc @space_cap]

@hybridherbst Whats about motion can the dino head turn to right side? Or the woman face make a big smile?

[sc @space_cap]

@ItsmeAjayKV Does this not depend on what the model was trained for? Maybe this was not a target of grok to train for? Maybe you need to use prompt like build next starship or a mars colonies?

[sc @space_cap]

@elonmusk It is realy fast evolving in my opinion regardingng architectures and also regarding fast reacting.
Let us think what happens if light or alternative processing hardware if needed, gets into this game. 🤯

[sc @space_cap]

@elonmusk Never spend energy into war there are more important challenges on earth and space!
We are one species on this planet, soon there will be others more intelligence ones. So let us be best acting parents to get for our kids the best future they can have!

[sc @space_cap]

@cb_doge 🖖

[sc @space_cap]

@elonmusk @cb_doge We need to think about:
How do we react at day X if an alien would arrive at earth and letˋs say it is a friendly alien.
During conversation we find out it is 1.3 times more intelligence and can super fast drive on 4 wheels. This is the first level we need to think about…

[sc @space_cap]

@CompleteSkeptic Jev sounds like the small brain for fast actions vs LLM.

Cognitive and Emotional Tasks fast:

Recent research shows that the cerebellum is also involved in language processes, thinking tasks, and the storage of emotional memories.
Faster needs to be Spinal Reflexes (Shortcut)

[sc @space_cap]

@elonmusk @grok @bot I think the Roadster comes back with some alien upgrades! Is landing zone prepared? With some alien acceleration and inference engine on board.

[sc @space_cap]

@elonmusk @farzyness Good things need time !

[sc @space_cap]

@MiaAI_lab Tetris can be build with small models like qwen3.8:27b and some stylesheets css is an add on top. But looks cool! So this can be done on smaller GPU’s as DGX Spark.

[sc @space_cap]

@beamnxw This fits also very well, with a different paper release in the recent days.

This paper was regarding agent how they loose the focus on target over time.

If you write up a plan or specification agents will stay focused especially for different agents and responsibilities team.

[sc @space_cap]

@rasbt All this different architectures wonderful explaind!
Feels like the future will by dynamic architecture where the agent get under genetic evolution algorithm climbing up in species levels of intelligence. How large does a model need to be to find the big steps upward fast? 1..27B

[sc @space_cap]

@JensenHuang @ChaseLochmiller @OpenAI And this all started by GPU like RTX 1050 with up to 4GB VRAM for 100 $ back in 2017 and paper “Attention is all you need”
Today we get RTX 5090 32GB VRAM for over 4500 $ rising.

[sc @space_cap]

@elonmusk @tobi @jimcsek @CathieDWood But wrong name, human intelligence is underhyped.

[sc @space_cap]

@sama At some point these huge models can also stup powerfull smaller local models with agentic team harness.
This will change how we live the next couple of years.
Sam it is very important to do this steps together and in open way, for our kinds and human species! No rush! Hope?

[sc @space_cap]

@MaxSumrall Can it forward roll like sonic ?

[sc @space_cap]

@busypussycat @StairwayToRetro Trust me this days come back but tools will be local llm. To have the huge ram and cpu/gpu processing under control. A farmer needs also big 🚜.