The AI Alliance launches Project Tapestry to build shared infrastructure for open and sovereign AI models in response to risks of Chinese open-weight systems dominating globally

The salvation is Project Tapestry https://thealliance.ai/projects/tapestry

@ylecun The open-source AI race will define tech leadership this decade. China's model velocity suggests they'll dominate by 2030 if the West doesn't act fast. Tapestry is a start, but we need more. TopTokens AI tracks these shifts daily.

@Dan_Jeffries1 Why an "American open source champion", why not look at the Human Genome Project as template (20 labs, 6 countries)

https://ngranati.substack.com/p/from-jaipur-to-deepseek-the-call?r=44hcrd&utm_campaign=post&utm_medium=web&triedRedirect=true

@Dan_Jeffries1 What happened to the Brave New World you assured us was coming? It was only a few months ago. Now there are just "consequences".

The technology is still nascent. It will become a commodity in time. For now, assembling training data is expensive, and competition is good. I say this as a huge proponent of personal computing. The justification for cloud data traps (like X) is far weaker than that for closed models, for many reasons.

@Dan_Jeffries1 Base model origin is a red herring. Power concentrates at fine-tuning and inference at the edge - not pretraining. That's where the real dependency forms, and where lock-in lands.

This isn't the social media race. No serious enterprise or government builds critical infrastructure on the most popular model they build on the most capable one. Anthropic has a fraction of ChatGPT's users but runs US classified missions, serves Netflix, Spotify, KPMG, and holds a $200M DoD contract. The world won't default to Chinese open models because they're free they'll choose whatever they trust with their most sensitive systems. Quality captures infrastructure. Popularity captures teenagers.

@Dan_Jeffries1 So, the rest of the world get the best deal? Good. let the west lost!

@Dan_Jeffries1 It’s healthy for everyone this way , America needs a wake up call .

@Dan_Jeffries1 Expect Google and Microsoft to lean into open weight hard.

Gemma4 is fantastic. Google’s coming around to the hardware game. Amazon might enter the fray soon too.

@Dan_Jeffries1 The West is becoming an imitation of the Soviet Bloc. The EU is even more laughable.

@Dan_Jeffries1 It didn't matter where Linux came from. In this case, it's also very early days - ten years from now, ask me what the plan is, once again.

@Dan_Jeffries1 It doesn't really matter if the open models are american or Chinese. The open models only cause damage to the country whose stock market is heavily dependent on ai monopolization, right now that's the USA. Open models from america will destroy the Ai mirage economy in USA as well

@Dan_Jeffries1 Elon has the opportunity to do this with grok but he won't because all his "open ai" talk is BS.

There's a credibility problem baked into the "national security" framing that doesn't get enough scrutiny.

Restricting Chinese open-weight models while having no viable Western open alternative just narrows the options for everyone outside the U.S. and accelerates exactly the outcome the restrictions are supposedly preventing.

Six billion people will use whatever works and is accessible. Policy that ignores that isn't serious policy.

@Dan_Jeffries1 How would you propose banning an open source model?

@Dan_Jeffries1 “The best open-source AI models now come from China. Researchers everywhere are using them”. — Yann LeCun.

@Dan_Jeffries1 AI is an output of energy. China's investment in energy production gives it an edge that the US will struggle to catch up with despite having better AI models. Chinese affordable open weights model will get the low market today and eventually the high end tomorrow

@ylecun TechBrunch season finale gonna be open-source founders and closed-model CEOs arguing in a café for 3 hours straight

@Dan_Jeffries1 RSI leading to ASI is what makes this time different. We are at the end of human contribution to software and governments don't like to share power.