correct, a purely open-source volunteer model is not going to solve this.
Maybe a licensing change for Open LLMs, like GPL did — where you give the model openly, but with the condition that the companies licensing it have to make all 1-bit human feedback that they've collected
@soumithchintala
-
GPL-Style Licensing for Open Source LLMs with Data Sharing
By
–
-
Open LLMs Need Coordinated Human Feedback Strategy
By
–
Open LLMs need to get organized and co-ordinated about sharing human feedback. It's the weakest link with Open LLMs right now. They don't have 100m+ people giving feedback like in the case of OpenAI/Anthropic/Bard.
They can always progress with a Terms-of-Service arbitrage, but -
Emergent Complexity in Neural Networks Through Gradient Descent
By
–
the non-linearity by itself isn't complex, something as simple as max(x, 0).
the system of two matmuls and a max in between, combined with gradient descent — that system becomes fairly complex. -
Understanding Two-Layer Perceptron Optimization Dynamics
By
–
the optimization dynamics of a two-layer perceptron are barely understood. that's just two matmuls and one pointwise non-linear function in between
-
AI and Brains: Reductionism in Intelligence Systems
By
–
AI is just matrix multiplications.
Brains are just biochemical interactions.
Must be simple. -
Synthetic Data Replacement Ethics and Regulatory Considerations
By
–
another caveat — if they basically replace the grandfathered stuff with "synthetic" data, which is "checked" to be non-overlapping, but covers the knowledge sufficiently.
-
AI Training Data: Addressing ToS Violations and Grandfathering Issues
By
–
easily will take the other side of the bet, unless they do it with big caveats such as only having this policy for newly sourced data (and grandfather in all their existing stuff, including general GenAI companies' rumored scraping of sketchy ToS violating stuff such as youtube)
-
Jiayq Launches New AI Inference Platform
By
–
.
@jiayq of caffe and pytorch fame just launched his AI Inference platform. check it out… https://
t.co/Nv9QX41Ga7 -
When does LLM regurgitation become original creative thought?
By
–
the origin of a human idea seems to be so poorly understood, that we claim both that
* LLMs are able to have creativity
* LLMs are unable to have creativity because they only regurgitate data at what point is it regurgitation and at what point does it become original thought? -
AI versus electricity: Comparing technological importance
By
–
uh, not me actually — maybe you just misattribute all those analogies to me — AI is not electricity and electricity is not AI. Both are cool things though and its not clear that AI is more important than electricity. The first person I saw compare AI to electricity is Andrew