> The major AI gatekeepers, with their powerful models, are already experiencing capacity and scale issues. This won't change unless the underlying technology (LLMs) undergoes a fundamental shift. As more and more things become AI-enabled, how dependent will we be on these gatekeepers and their computing capacity? And how much will they charge us for prioritised access to these resources? And we haven't really gotten to the wearable devices stage yet.
The scale issue isn't the LLM provider, it's the power grid. Worldwide, 250 W/capita. Your body is 100 W and you have a duty cycle of 25% thanks to the 8 hour work day and having weekends, so in practice some hypothetical AI trying to replace everyone in their workplaces today would need to be more energy efficient than the human body.
Even with the extraordinarily rapid roll-out of PV, I don't expect this to be able to be one-for-one replacement for all human workers before 2032, even if the best SOTA model was good enough to do so (and they're not, they've still got too many weak spots for that).
This also applies to open-weights models, which are already good enough to be useful even when SOTA private models are better.
> You could argue that we already send a lot of data to public clouds. However, there was no economically viable way for cloud vendors to read, interpret, and reuse my data — my intellectual property and private information. With more and more companies forcing AI capabilities on us, it's often unclear who runs those models and who receives the data and what is really happening to the data.
I dispute that it was not already a problem, due to the GDPR consent popups often asking to share my browsing behaviour with more "trusted partners" than there were pupils in my secondary school.
But I agree that the aggregation of power and centralisation of data is a pertinent risk.