Hugging Face CEO Clément Delangue argues concentration of AI power and wealth is the industry's greatest risk
Delangue advocated for open-source models to counter consolidation
Positive users back the Hugging Face CEO's warning on AI power concentration by urging open source to democratize access and ensure provenance, while negative users argue training costs and other risks matter more than open code.
Most Activity
The fight continues.
Concentration of power, capabilities and economic wealth is the biggest risk in AI. We need open science and open-source more than ever!
@Plinz Most of the other risks are, in my view, either much less likely, or made more likely and more damaging by extreme concentration of power, capabilities and wealth
@ClementDelangue I hate to tell you but there are even bigger risks than someone having a better AI model than you
@ClementDelangue I hate to tell you but there are even bigger risks than someone having a better AI model than you
Concentration of power, capabilities and economic wealth is the biggest risk in AI. We need open science and open-source more than ever!
@ClementDelangue "The only safe models are the ones that make us tons of money and put everyone else out of work" is quite the message to rally behind...
This is not one that we can let go quietly.
@ClementDelangue Yes.
Concentration of power, capabilities and economic wealth is the biggest risk in AI. We need open science and open-source more than ever!
@ClementDelangue open source should win
@ClementDelangue Yep
@ClementDelangue Couldn't agree more. AI is built on the collective wisdom of humanity over millennia—it belongs to all of us. Any move to privatize AI is purely selfish and narrow-minded, and it’s a direct threat to social order and fairness.
@ClementDelangue the conviction is right but "open source" has become the thing everyone agrees on while disagreeing about everything underneath it - model weights without compute access, data provenance, or governance is a flag without a country
I’d love to see Hugging Face add a clause requiring that any use of its datasets as well as any dataset uploaded to its platform must be subject to a mandatory reciprocal open research license, regardless of the company involved. This is probably the only real way to stop these players from undermining open research and the community.
@ClementDelangue let's get the boys together and finally train that 10T we have always been talking about
@ClementDelangue the polymarket board showing anthropic at 91% would like a word
real!!
Most problems, because of the unfortunate way monopolies work, is that you need like a few solutions for there to be that market that efficiently starts pricing the products close to how much they actually cost.
So we can't just solve aging once at one company. As it will be price gouged.
We need many companies to all solve aging in biology. So it can be an affordable solution for everyone
That includes giving the startups a chance, which Anthropic is not doing right now.
@ClementDelangue i need gpus/funding to train a text autoencoder on all accepted papers in existence from top ml conferences.
outcome: new opensource non-LLM genAI.
fund this.
@ClementDelangue The tech is exciting. The concentration of power around it is the part that keeps me thinking🤔
@ClementDelangue Do you think large-scale distributed training and inference projects will gain traction?
@ClementDelangue That’s definitely a big risk but I don’t think it’s the biggest. What about extinction?
@Plinz @ClementDelangue funny how there are people who are acutely aware of that, and people who are not even remotely aware of that
Open source is not just ideology here.
It is provenance.
When the model, weights, routing, safety interventions, and access tiers are closed, users cannot tell whether failure came from:
bad idea bad implementation model limitation hidden provider intervention commercial boundary enforcement
That ambiguity concentrates power.
Open science means the experiment can be inspected. Open source means the tool can be audited.
Without both, AI becomes rented intelligence with invisible governors.💰💰💰💰🦜
@ClementDelangue on it clem sensei 🫡 don't worry