yawn

Why pay for really expensive models when you can get almost the same performance for a fraction of the cost?

The future is not about the most powerful models necessarily. There will be cases, and there will be a development in intelligence going forward.

My prediction is that the future of AI (LLM++) will be harnesses and tools, not the absolute best inference.

But Amodei says open models are bad.

Fucking “Hugging Face”?