Within the final 60 years expertise has advanced at such an exponentially quick price that we at the moment are repeatedly conversing with AI based mostly chatbots, and that very same OpenAI expertise has been put right into a humanoid robotic. It’s actually wonderful to see this fast improvement.
Continued development of AI improvement faces quite a few challenges. Considered one of these is computing structure. Because it was first described in 1945, the von Neumann structure has been the inspiration for many computing. On this structure, directions and knowledge are saved collectively in reminiscence and talk by way of a shared bus to the CPU. This has enabled many many years of steady technological development.
Nevertheless, there are bottlenecks created by such an structure, when it comes to bandwidth, latency, energy consumption, and safety, to call a number of. For continued AI improvement, we are able to’t simply make brute power changes to this structure. What’s wanted is an evolution to a brand new computing paradigm that bypasses the bottlenecks inherent within the conventional von Neumann structure and extra exactly mimics the system is attempting to mimic: the human mind.
To attain this, reminiscence have to be nearer to the compute engine for higher effectivity and energy consumption. Even higher, computation needs to be achieved straight inside the reminiscence itself. This paradigm change requires new expertise, and ReRAM (or RRAM) is among the many most promising candidates for future in-memory computing architectures.
RelatedPosts
