Keep pulling the thread on Clem Delangue.
The Hugging Face pivot was sparked when co-founder Thomas Wolf ported Google's newly released BERT model from TensorFlow to PyTorch over a weekend.
Hugging Face is the number one platform for AI builders.
Hugging Face has over 5 million AI builders using its platform daily.
Users have shared over 3 million models, datasets, and applications on the Hugging Face platform.
A new model, dataset, or application is built on the Hugging Face platform every 10 seconds.
A future where AGI is controlled by a single company or organization represents the highest risk, whereas decentralized access to the technology creates a safer future.
The "Lean Startup" methodology is not effective for AI companies because they are often founded by scientists and require heavy capital investment before generating returns.
Hugging Face has raised over $500 million in funding over 7 years.
Hugging Face is a profitable company and has spent less than half of the capital it has raised.
The current trend of AI companies building on top of APIs is a transitional phase, not the optimal long-term approach to building AI products.
Eventually, all tech companies will become AI companies and will build, optimize, and fine-tune their own models for their specific use cases.
The future of AI will involve a proliferation of smaller, more efficient, and cheaper models optimized for specific domains, leading to a world with as many models as there are code repositories.