Of course it is, the playing field is stacked with companies leeching off it - massive profit margins for TSMC, Nvidia, RAM manufacturers, AWS/Azure/Google Cloud, and a 30% fee for using your smartphone.
Ask yourself, why is an exec from Nvidia saying this right now?
Why is he declaring this now as a problem, when they know that they want the exact opposite to be true?
Who is this Nvidia exec really talking to?
He is talking to the ones invested in the AI data center build out: OpenAI, CoreWeave, Oracle, Softbank and Microsoft and many other investors. But he is really saying this to them:
Translation: The problem around paying workers still exists and that is cheaper than running AI agents for now. Once we have efficient AI models both cloud and local, we bet that due to Jevons paradox, as the cost per token goes down and as soon as the data center buildout is complete, we can meet the demand to run millions of AI agents cheaply and good enough to stop paying human workers and to eventually replace them.
Discussions (88 points, yesterday, 61 comments) https://news.ycombinator.com/item?id=47918009
(35 points, 6 hours ago, 7 comments) https://news.ycombinator.com/item?id=47936209
Of course it is, the playing field is stacked with companies leeching off it - massive profit margins for TSMC, Nvidia, RAM manufacturers, AWS/Azure/Google Cloud, and a 30% fee for using your smartphone.
The cost comparison is misleading. AI doesn't replace human judgment - it replaces human processing. Different problem entirely.
AI is parallelizable, and there's no shortage of silicon, ultimately it will come down to energy price
For now.
Ask yourself, why is an exec from Nvidia saying this right now?
Why is he declaring this now as a problem, when they know that they want the exact opposite to be true?
Who is this Nvidia exec really talking to?
He is talking to the ones invested in the AI data center build out: OpenAI, CoreWeave, Oracle, Softbank and Microsoft and many other investors. But he is really saying this to them:
Translation: The problem around paying workers still exists and that is cheaper than running AI agents for now. Once we have efficient AI models both cloud and local, we bet that due to Jevons paradox, as the cost per token goes down and as soon as the data center buildout is complete, we can meet the demand to run millions of AI agents cheaply and good enough to stop paying human workers and to eventually replace them.
Do not fall for it.
No shit. Everyone called it but you guys, because you didn't want to.