Thanks for a very interesting post, but there is one possible flaw: comparing the deployment of hardware (electrification) with software (LLMs) is not fair. Their dynamics are profoundly different. Compare like-systems, please.
> But we’ve only scratched the surface about AI diffusion
One area that would be interesting to analyse is the diffusion of AI through open source software and hubs like Hugging Face (disclaimer: I work there). Although smaller in scale than consumer apps like ChatGPT, the adoption of local LLMs and tooling like llama.cpp is likely exhibiting similar rates of growth that a token-based analysis of APIs is blind to.
Great post! Would it be useful to look into how fast the economic value from AI as a fraction of the gross world product has grown in comparison with other tecnologies?
Thanks for a very interesting post, but there is one possible flaw: comparing the deployment of hardware (electrification) with software (LLMs) is not fair. Their dynamics are profoundly different. Compare like-systems, please.
Excellent post!
> But we’ve only scratched the surface about AI diffusion
One area that would be interesting to analyse is the diffusion of AI through open source software and hubs like Hugging Face (disclaimer: I work there). Although smaller in scale than consumer apps like ChatGPT, the adoption of local LLMs and tooling like llama.cpp is likely exhibiting similar rates of growth that a token-based analysis of APIs is blind to.
Great post! Would it be useful to look into how fast the economic value from AI as a fraction of the gross world product has grown in comparison with other tecnologies?