Sabitlenmiş Tweet
UAE will recover at a speed that will shock the whole world.
Do not underestimate the creativity, financial/diplomatic power, and sheer will of MBZ, Al Maktoum, and their government.
They have faith, so I do. #UNITED 🇦🇪
English
Armin - Hub.xyz
12.4K posts
Armin - Hub.xyz
@xarmin
@hubxyz co founder @gensynai @doublezero seed @nolimithodl @q42_co @superscrypt LP
Globe Katılım Eylül 2022
4K Takip Edilen7.8K Takipçiler
AI models are eating faster than the internet can grow
Epoch AI published one of the most important papers in the AI industry. They estimated the total stock of quality human-generated text at ~300 trillion tokens, and projected an 80% probability that frontier models exhaust it between 2026 and 2032.
We are inside that window now.
But the real crisis is worse than the paper predicted.
The supply isn't just finite. It's shrinking. Over 74% of newly created web pages now contain AI-generated text. The internet is being contaminated by the very models it trained. Every generation of AI pollutes the training data for the next one.
The industry tried synthetic data as a fix. The research killed that idea. Models trained on their own outputs degrade. They lose the nuance and variability that made human data valuable. Without real human data as an anchor, quality collapses.
Meanwhile, training costs keep climbing. GPT-4 cost ~$40M. The current generation is approaching $1B. The next: $10B+. Every dollar spent scaling compute accelerates the depletion of the data these models depend on.
Ilya Sutskever said it at NeurIPS: "We've achieved peak data. There's only one internet."
And that's just text.
The industry is shifting multimodal. Models now need real-world audio across hundreds of languages. Real images from real environments. Real video from natural conditions. That data doesn't live on the internet. It hasn't been collected yet.
The race for compute is solved by capital. The race for data has no equivalent solution.
You can't manufacture more internet. You can't synthesize what doesn't exist. And you can't scrape what was never captured.
This is what @hubxyz is building.
The data AI needs next doesn't exist online.
It exists in the real world.
And someone has to go collect it
Epoch AI@EpochAIResearch
Are we running out of data to train language models? State-of-the-art LLMs use datasets with tens of trillions of words, and use 2-3x more per year. Our new ICML paper estimates when we might exhaust all text data on the internet. 1/12
English
AI companies spend billions acquiring training data
5 billion people carry cameras and microphones in their pockets
Nobody connected the two
Until now
We built a new platform where anyone with a phone can collect real-world data for AI and earn from it.
A student in Manila recording how Tagalog sounds on a crowded bus.
A mother in Lagos photographing the objects on her kitchen table.
A woman in Bogotá capturing how afternoon light falls through her window.
Real images. Real audio. Real environments.
From 190+ countries.
Data that no AI company can buy from any provider today because it does not exist yet.
Hub already delivers structured training data to companies building the next generation of AI models.
Now we are opening the supply side to the world.
The people left behind by the economy just became its most valuable participants.
The first tasks are live.
The first contributors are already earning inside our Discord.
This is not a test. This is the beginning.
The full public launch is coming.
English
@hubxyz Data that didn't exist 48 hours ago, full launch is next
English
@thatismyquant There is indeed still no second best.
UAE’s pmf is not going away 😉
English
UAE will recover at a speed that will shock the whole world.
Do not underestimate the creativity, financial/diplomatic power, and sheer will of MBZ, Al Maktoum, and their government.
They have faith, so I do. #UNITED 🇦🇪
English
It’s increasingly clear that the UAE will emerge stronger than ever before from these brutal and unwarranted attacks.
They have:
- Protected the population (mostly non Emirati citizens)
- Looked after those least fortunate.
- Maintained calm and composure.
- Opened all services as soon as practicable.
- Avoided unnecessary escalation.
These are the grownups in the room.
Little Sparta indeed! 🇦🇪 🇦🇪 🇦🇪
English
Exactly this is why Dubai will win endlessly going forward
Living rent free in everyone’s mind
Objectively superior societal model, globally
Check back on this tweet in a year I dare ya
أحمد خليفة@_A_khalifa
If Dubai doesn’t vibe with you, no one’s forcing you to yap about it! And if you secretly hate Dubai that much, maybe skip the fake news fanfiction next time? Simple as that. Chill, habibi. 😏
English
Dubai’s success seems to bother people who’ve never built anything themselves
English
Every Iranian drone you see the UAE destroying is not just one less threat in our skies. It is one LESS DRONE that WILL REACH EUROPE'S borders, one less weapon that could have killed Ukrainians, and one less instrument of terror in the hands of a Islamic regime that exports chaos. When the UAE intercepts them, it is not only defending its own people. It is quietly saving lives far beyond its borders.
English
Say hello to TesterArmy (YC X26)!
Our mission is to change how AI-native teams tests their apps.
→ Tests every Pull Request
→ One-click integration with GitHub and Vercel ▲
→ Explores pages, clicks buttons, fills forms
→ Reports back actionable feedback
English
Armin - Hub.xyz retweetledi
Dario Amodei listed 7 factors that determine how capable AI becomes
Compute. Data quantity. Data quality. Training duration. Objective function. Two normalization terms.
The industry is pouring $100B+ into factor one. Factors two and three have no infrastructure at scale.
No distributed pipeline. No real-time supply chain.
Hub knows the true bottleneck. Our next step is fixing it.
Rohan Paul@rohanpaul_ai
Anthropic CEO Dario Amodei: AI Progress isn’t magic, it’s just compute, data, and training. "All the cleverness, all the techniques, all of the “we need a new method,” doesn’t matter very much. There are only a few things that matter, and I listed seven of them. One is how much raw compute you have. Two is the quantity of data you have. Three is the quality and distribution of the data, meaning it needs to be broad. Four is how long you train for. Five is an objective function that can scale to the moon. Pre-training is one. Another is an RL-style objective: you have a goal and you go reach it. That includes more objective rewards (like math and coding), and more subjective rewards (like RL from human feedback, or higher-order versions of that). Six and seven are about normalization/conditioning: keeping numerical stability so the huge blob of compute flows cleanly instead of blowing up or getting weird." --- From 'Dwarkesh Patel' YT Channel (link in comment)
English
Armin - Hub.xyz retweetledi
Armin - Hub.xyz retweetledi
NVIDIA just removed the biggest friction point in Voice AI 🤯
They've open-sourced PersonaPlex-7B, a full-duplex conversational model that can listen and speak at the same time.
Instead of waiting for you to finish talking, it uses a dual-stream architecture to process and respond in real-time.
100% Open-Source and Free.
English
