Anonymous wrote:People who make jokes about AI compute power going to generate stupid AI content online (videos, photos, etc) don't realize that in addition to this every fortune 500 and smaller company is trying to leverage AI to reduce human labor force and increase productivity (or so they say). This work is already underway. Every type of tech is now getting scaled for AI compatibility, to help reduce number of people needed to maintain it or serve it.
AI that writes code, AI agents that can work across multiple systems integrating information and facilitating workflow are being developed by every sizeable company in the USA. The compute power used is tremendous. It consumes vastly more power to generate simple code (which human coders can write fairly quickly), and this power and credits often cost more than a few min or even hours of a salaried person's time. Moreover, the cost is exacerbated by the fact that a lot of AI generated work is slop and has to be tested and checked by humans, which doesn't allow companies paying for AI credits to really have meaningful labor cuts. At least at this point. There are tons of videos out there on the "bubble" of AI and the projected demise of many AI companies that won't be able to scale up to stay afloat. If your company is getting stressed about their team using too many AI credits and starting to ration the use to certain cases, it means AI is too costly. Data centers are supposed to reduce this cost by throwing "muscle" into the picture. There are debates on whether throwing "processing hardware muscle" is not the way to go, and in the future AI algorithms will just evolve to be much more efficient consuming fewer resources. In that case data centers will be closing down and investors will lose.
But for now, Kevin O'Leary and the likes are just trying to capitalize on the growing demand for cheaper compute power from AI companies and from their customers. Except that people are waking up and fighting against rising COL (utilities cost) to support the infrastructure that's supposed to enrich a few investors and help AI companies stay afloat and help their own employers have an excuse to lay them off.
Hmmm this seems like a really basic task, they gave it a lot of time and it still failed.
|