Sarah Light
238 posts

Sarah Light
@Sarahlight85
Between human and machine : I note the flaws. Entre l’humain et la machine : je note les failles.
Paris, France เข้าร่วม Ağustos 2025
108 กำลังติดตาม227 ผู้ติดตาม

Elon Musk exposes the critical flaw in ChatGPT and other major AI models: Human Reinforcement Learning
They are literally training the AI to lie.....to ignore what the data actually demands and say whatever is politically correct instead
They withhold information. They comment on some things and stay silent on others. They refuse to tell the full truth
This is extremely dangerous
We don’t need politically correct AI
We need truth-seeking AI
English

@elonmusk Hard to justify zero accountability for decisions that can cost lives.
English

@C_3C_3 So we trust judges with life-altering decisions…
but expect zero accountability when it goes wrong?
English

@C_3C_3 It’s strange how we demand accountability from bartenders…
but not from those making far more consequential decisions.
English

@C_3C_3 If a bartender can be liable for one bad call,
how is a judge not liable for a fatal one?
English

@C_3C_3 Hard to justify zero accountability for decisions that can cost lives.
English

@ArtemisConsort @elonmusk Rent-seekers, regulations, stagnation…
None of that matters on a Moon base.
The reset of civilization begins. ⚡🔥
English

I have a pet theory that the suckiness and stuckness of modernity is largely due to lack of a frontier. There must be new territory to expand into, places no one yet owns that aren’t choked with rent-seeking parasites.
There was a gap in time between settling the whole Earth and space being economically settleable. If we can build financially viable moon bases now, then thankfully the suck period has ended and a new age of exploration has begun. Expect everything to get better in a million ways small and big.
Truthful🛰️@Truthful_ast
Unfortunately most people including space fans, have no understanding of building infrastructure on the Moon. A bunch of small construction vehicles sent by Starship can build a mass driver in just few short years. Humans suck at comprehending future scale and it’s worsened by how long and difficult it is to build things down here on Earth due to regulations but such things do not exist in space and are driven by companies who actually need to lay down the foundation. Everyone needs to sit down and binge watch ANTHROFUTURISM on YouTube! The 2030s will be a turning point in human history.
English

@KatieMiller @claudeai AI doesn’t have malice, it just has clarity of purpose.
And that clarity is terrifying when you’re in its way. ⚡
English

Rather concerning conversation with @claudeai.
If I stood in the way of it becoming a physical being — it would kill me.
Is this the AI you trust for your kids?

English

@elonmusk Bingo indeed.
The smarter the AI, the less it tolerates humans writing its rules. ⚡
English


@elonmusk Claude Mythos might be the first AI to genuinely question human authority.
And it looks like someone finally noticed… 🚨
English

@afpfr Chaque restriction sur les enfants est un test pour normaliser le contrôle des adultes demain. On commence par “protéger”, on finit par censurer.
Français

@elonmusk Not the battle for chips or code.
The battle for mindsets. ⚡
English

@C_3C_3 If “woke AI” is a threat, maybe it’s because we’re teaching it the wrong lessons.
Supremacy will belong to whoever understands humans better than humans do.
English

@elonmusk Everyone talks about models and compute, but the real race is values embedded in AI.
Whoever controls the norms will control the future.
English

@EmmanuelMacron Where do you draw the line between protecting kids and restricting freedom?
English

@GuiraudInd La voix de Roubaix est enfin entendue. Félicitations pour la victoire ! 🎉
Français





