4DS - Anything but Charting, page-7772

Hash,

I reckon that building the most performing non volatil memory is the main task of 4DS.
They most probably want to stay on their track until they reach their target.

But, at the same time, there is an AI revolution going on.
This revolution will change almost everything if we are to believe ARM. Really, really huge...

https://www.arm.com/solutions/artificial-intelligence

And NVIDIA seems to believe it, as they want to spend 40 billions $ to buy ARM to build the most powerfull company of the AI era.

https://www.theverge.com/2020/9/14/21435890/nvidia-arm-acquisition-40-billion-ai-cloud-edge-why

... And not only for edge computing but also for data centers.

"..Nvidia, for a start, can potentially leverage the efficiency of Arm’s CPUs to bring down power consumption in its data centers. Although edge AI is a big part of the field’s future, data centers aren’t going anywhere either, and the energy costs for running them are mammoth. Nvidia CEO Huang himself suggested this would be a focus of the acquisition when he noted on this morning’s call that “energy efficiency is the single most important thing when it comes to computing going forward.” Arm, meanwhile, should benefit hugely from Nvidia’s AI expertise and resources. Nvidia has an impressive reach in the world of machine learning, spanning fields like robotics, self-driving cars, and medical imagining, as well as consistently publishing interesting and novel ML research. It also simply has the size to accelerate any R&D efforts Arm needs in order to push its AI capabilities further. "

As I said in a previous post, ReRam is believed to be the best technology for AI.
I finaly found why, reading this abstract of an academic paper by Sparsh Mittal.

https://www.academia.edu/36504841/A_Survey_of_ReRAM_based_Architectures_for_Processing_in_memory_and_Neural_Networks

"...By comparison [ with others NVM], ReRAM is a non-volatile memory with near-zero leakage energy and high density.The ReRAM state reflects the current passed through it in the history and this is very usefulfor modeling the synaptic weights of neurological synapses and implementing neural network(NN) architectures [9–11]. In addition, ReRAM supports operations such as analog matrix-vectormultiplication (MVM), search and bitwise operations within memory which facilities energy-efficientaccelerator design.
For example, since the CONV (convolution) operation in convolutional neural networks (CNNs) involves MVM, and CONV layers account for more than 95% of computations inCNNs, ReRAM-based processing engine can boost the efficiency of CNNs significantly [12,13]. Thesefactors have motivated researchers to implement a variety of ML/NN architectures on ReRAM, suchas multi-layer perceptron [14,15], CNN [16–18], tensorized NN [19] and auto-associative memory [14]"

But
"Use of ReRAM for designing neuromorphic computing systems (NCSs), however, also presents challenges. For example, ReRAM has several reliability issues, such as limited write endurance, resistance drift, susceptibility to process variation (PV), etc. [20–23]. "

It's looks like the 4DS non filamentary ReRam adress those challenges.

What do all that mean at this point of the 4DS journey ?

I know that if you want to reach your goal, you must stay focused on what you decided first.
But we are in semiconductor world. A world of ever expanding possibilies (And, for poor investors, a world of infinit iterations).
And quit often scientific discoveries are made by accident.

So I think, we can compare the 4DS journey with Christopher Colombus voyage. The first of all modern ventures.
Same anguish, same doubts, same willings to throw the captain over the board and to turn back. All these sufferings to discover a new road to ... India !
They finaly discovered America.

Could IA be the America of 4DS ?