'TurboQuant' is a data compression algorithm that solves in-memory bottlenecks and maximizes AI efficiency. The algorithm reduces memory usage while maintaining accuracy. It was also mentioned that memory usage required for specific tasks was reduced by more than 6 times.
However, there are indications that such concerns are excessive. This is because it is difficult to find cases where memory demand decreases compared to before when new technologies that improve resource utilization emerge.
There are also forecasts that in the long term, it will act in the direction of increasing memory demand. The explanation is that it will accelerate AI popularization and instead cause memory demand to surge. Morgan Stanley analyzed, "It will not reduce total overall memory demand, but will be a catalyst for expanding the AI market pie itself." While doing so, it recommended, "Rather, it should be seen as a buying opportunity."
===
They said it reduced it by 6 times for specific tasks, but is this really something to react so sensitively to..
Is this a bargain sale!?
